US20080019162A1

US20080019162A1 - Non-volatile semiconductor storage device

Info

Publication number: US20080019162A1
Application number: US11/758,108
Authority: US
Inventors: Taku Ogura; Masaaki Mihara; Yoshiki Kawajiri
Original assignee: Genusion Inc
Current assignee: Genusion Inc
Priority date: 2006-07-21
Filing date: 2007-06-05
Publication date: 2008-01-24
Also published as: US20090073740A1; US7969780B2

Abstract

This non-volatile semiconductor storage device includes a flip-flop in which two inverters, each consisting of a load transistor and a storage transistor connected in series, are cross-connected; and two gate transistors, each respectively connected to a node of the flip-flop on a side thereof. The storage transistors of the inverters are constituted by storage transistors which can be threshold voltage controlled by injection of electrons into the neighborhood of their gates. This non-volatile semiconductor storage device further includes two bit lines, each of which is connected to a respective one of the two gate transistors; a word line which is connected to both of the gate electrodes of the two gate transistors; a first voltage supply line which is connected to the sources of the storage transistors of the inverters; and a second voltage supply line which is connected to the sources of the load transistors of the inverters.

Description

CROSS REFERENCE

This Nonprovisional application claims priority under 35 U.S.C. § 119(a) on Patent Application No. 2006-199673 filed in Japan on Jul. 21, 2006, the entire contents of which are hereby incorporated by reference.

BACKGROUND OF THE INVENTION

The present invention relates to a non-volatile semiconductor storage device which can be electrically erased and written, to a state determination method for such a semiconductor storage device, and to a semiconductor integrated circuit device incorporating such a semiconductor storage device.
Along with increase of internal SRAM capacity, the necessity has increased for providing redundancy, for performing individual tuning after making boards such as LCD drivers and the like, and for low cost fuses which are being demanded along with the great increase in various types of application which involve personal identification information (ID codes, encryption and decryption keys, numbers of IC cards, and the like).
In the prior art, as fuse elements which can be made with a standard CMOS process, there is a known one in which a polysilicon or metal wiring layer is blown out by a laser or an electrical current, a known one in which an insulating gate layer or the like is destroyed by voltage, and the like. However, such a fuse which employs such a blowing out or insulation destruction method or the like is not suitable for the application described above, since it can only be programmed once.
On the other hand, in the case of using a non-volatile element of the floating gate type which can be manufactured by a CMOS process, although it is possible to implement a fuse which can be electrically erased and written, introduction of a special process for making the transistors non-volatile, such with a prior art type flash memory, is not appropriate from the point of view of cost. Furthermore, with a floating gate type element made with a standard CMOS process, there has been the problem that the data storage characteristics deteriorate as the insulating layer becomes thinner along with increase of the integration scale.
Thus in, for example, U.S. Pat. No. 6,518,614, Japanese Laid-Open Patent Publication 2004-56095, and Japanese Laid-Open Patent Publication 2005-353106, there are disclosed a non-volatile storage device which can be manufactured by a standard CMOS process, and a non-volatile storage device which does not have any special floating gates.
FIG. 1 is a figure showing the memory cell structure of a non-volatile storage device which has been manufactured by a standard CMOS process, as disclosed in Japanese Laid-Open Patent Publication 2005-353106. Fundamentally, this consists of N-type MOS transistors MCN1 and MCN2 which constitute a non-volatile data storage unit, and a flip-flop unit (MN3, MN4, MP1, and MP2) of a static latch type which takes the output nodes T and B of the non-volatile data storage unit as differential inputs. In this flip-flop unit, normal reading out and writing SRAM operation is performed. Furthermore, it is possible to reload the information of the non-volatile data storage unit and to store data in the flip-flop unit.
FIG. 2 show the data setting method. This data setting method is a method in which the data is determined by the voltage difference between the threshold values of the two N-type MOS transistors MCN1 and MCN2. In the initial state (FIG. 2A) before writing data, the N-type MOS transistors MCN1 and MCN2 both have the threshold voltage Vth, and, in this state, the output data of the flip-flop is indeterminate. In order to determine the data, first, writing of the data to “0” is performed (FIG. 2B). This writing is implemented by raising the threshold voltage on the side of MCN1 to Vth1 (where Vth1>Vth0). With this structure, it is not possible to perform erasure (reduction of Vth). Because of this, subsequent writing of the data “1” from the data “0” state is implemented by raising the threshold voltage on the side of MCN2 to Vth2 (where Vth2>Vth1) (FIG. 2C).
FIG. 3 shows the method for changing the threshold voltages of the N-type MOS transistors of the above described non-volatile data storage unit. In FIG. 3, the data “0” case is shown, i.e. the case in which the threshold voltage on the side of MCN1 is raised. Fundamentally, positive advantage is taken of the deterioration characteristic of an N-type MOS transistor due to hot carriers. In other words, it is arranged to set to 0 V the source potential of that transistor MCN1 whose threshold voltage it is desired to raise, to set its gate potential (MLW) to 2.5 V, and to set its drain potential (the node T) to 5V, thereby raising its threshold voltage in the vicinity of its drain terminal by the hot carrier injection phenomenon. At this time, in order for the potential of the bit line BLT to be sufficiently supplied to the node T, it is necessary to raise the potential of BLT to 5 V, and to raise the potential of the word line WL of the flip-flop unit to 7V. As for that transistor MCN2 whose threshold voltage it is not desired to raise, by setting the potential of BLB to 0 V, it is arranged for hot carrier injection not to take place in the vicinity of the drain terminal of that transistor MCN2. And, when writing the data “1”, since the threshold voltage on the side of MCN2 is be raised, BLT is set to 0 V and BLB is set to 5 V. The other conditions are the same as when writing the data “0”.
FIG. 4 shows the method for transferring data transfer from the non-volatile data storage unit to the flip-flop unit. This figure shows the data transfer method in the data “0” case, in other words when the threshold voltage Vth1 of MCN1 is higher than the threshold voltage Vth0 of MCN2. In the state of the flip-flop unit in which the word line WL=0 V and a restore control signal RESTORE=0 V, by lowering an equalization control signal ZEQ from Vcc to 0 V at the time instant T0, the node T and the node B are equalized at the same potential. This equalization operation ends at the time instant T1. And MLW, which is the gate potential of the transistors MCN1 and MCN2, is gradually raised from the time instant T2. When this is done, MCN2 is turned ON first, being that transistor whose threshold voltage is the lower, and the potential at the node B is pulled down. Although MCN1 also goes ON after some time, finally the latch is established in the state in which the node B on the side of the transistor MCN2 whose threshold voltage is the lower has reached 0 V, while the node T on the side of the transistor MCN1 has reached Vcc. And the rising of MLW is completed at the time instant T3. By raising RESTORE from 0 V to Vcc at the time instant T4, the latching of the flip-flop unit is activated, and the data is held in a stable manner. Finally, at the time instant T5, MLW reaches 0 V and this process terminates.
While the operation of such a memory cell is as described above, there are the following problems with this structure.
(1) The margin of threshold voltage difference is small.
Vth1−Vth0 corresponds to this threshold voltage difference margin in the case of the data “0”, and Vth2−Vth1 in the case of the data “1”. To some extent, there is an upper limit value Vth_max for the amount of change of the threshold voltage due to the hot carrier injection phenomenon. If the margin is allocated equally between reading out the data “0” and the data “1”, then, if it is supposed that rewriting is performed once, the margin for each becomes (Vth_max−Vth0)/2. And, if it is supposed that rewriting is performed N times, then taking the maximum value of Vth_control as Vth_max, it must be divided into 2N portions, and the margin of each of the data “0” and the data “1” becomes (Vth_max−Vth0)/2N, so that the margin becomes yet smaller.
(2) As operating voltages when writing data into this non-volatile data storage unit, it is necessary to apply high voltages (7 V and 5 V) to the word line WL and to the bit lines BLT and BLB for each memory cell which it is necessary to control. This means that it is necessary to use high withstand voltage transistors for the drivers which drive the word lines and the bit lines, and for the column selection transistors for selecting the bit lines. Since the performance with high withstand voltage transistors, which are optimized for high voltage, is deteriorated when operating at Vcc=1.8 V as during normal reading out operation, accordingly the problem occurs that this entails access delay. And increasing the size of the transistors in order to increase the current drive capability causes the accompanying problem that the chip area becomes large.
One object of the present invention is to provide a rewritable non-volatile semiconductor storage device, with which it is possible to make the reading out margin large.
A further object of the present invention is to provide a rewritable non-volatile semiconductor storage device, with which it is possible to perform control of the word lines and the bit lines at the Vcc level.
Other objects of the present invention are to provide a state determination method for such a semiconductor storage device, and to provide a semiconductor integrated circuit device incorporating such a semiconductor storage device.

SUMMARY OF THE INVENTION

The non-volatile semiconductor storage device of the present invention includes a flip-flop in which two inverters, each consisting of a load transistor and a storage transistor connected in series, are cross-connected; and two gate transistors, each respectively connected to a node of said flip-flop on a side thereof.
Furthermore, the non-volatile semiconductor storage device of the present invention includes two bit lines, each of which is connected to a respective one of said two gate transistors, and a word line which is connected to both of the gate electrodes of said two gate transistors. These two bit lines and this word line are controlled to a voltage between operating power supply voltage and ground voltage.
Moreover, the non-volatile semiconductor storage device of the present invention includes a first voltage supply line which is connected to the sources of said storage transistors of said inverters, and a second voltage supply line which is connected to the sources of said load transistors of said inverters. A predetermined first voltage is supplied to the first voltage supply line during writing and during erasure. And a predetermined second voltage is supplied to the second voltage supply line during writing. This predetermined first voltage and this predetermined second voltage may be, for example, high voltages greater than or equal to the operating power supply voltage.
These storage transistors are constituted by storage transistors which can be threshold voltage controlled by injection of electrons into the neighborhood of their gates. In the non-volatile semiconductor storage device described above, the writing and erasure of data are implemented by controlling the threshold voltages of the storage transistors. These threshold voltages are provided via the first voltage supply line and the second voltage supply line which are connected to the two ends of the two inverters (the sources of the load transistors and the sources of the storage transistors).
Accordingly, when performing writing by elevating the threshold voltage of a first one of the storage transistors, and when performing erasure by lowering the threshold voltage of this first one of the storage transistors, this may be performed by the word lines and the bit lines operating at the level of the operating power supply voltage (Vcc). For this, in the above described non-volatile semiconductor storage device which makes it possible for peripheral circuitry to be more compact and be read out at higher speed, each of the storage transistors further includes an insulation layer side spacer which is formed at a side portion of its gate electrode; and a low impurity density region which is formed at a border portion of its drain.
During writing, said first voltage supply line applies said first voltage to the sources of said storage transistors, and said second voltage supply line applies said second voltage to the gates of said storage transistors via the sources of said load transistors; and thereby information is written into said storage transistors by channel hot electrons being injected into said insulation layer side spacer.
And, during erasure, said first voltage supply line applies said first voltage to the sources of said storage transistors; and thereby erasure of information which is stored in said storage transistors is performed by an avalanche of hot holes being injected into said insulation layer side spacer.
The above described first voltage and second voltage are high voltages which are greater than Vcc.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a figure showing the structure of a prior art memory cell;

FIGS. 2A to 2C are figures for explanation of the data setting method and the reading out margin for this prior art memory cell;

FIG. 3 is a figure showing the write voltage application condition for writing to a non-volatile data storage unit of this prior art memory cell;

FIG. 4 is a figure for explanation of a method of data transfer from a non-volatile data storage unit to a flip-flop unit of the above described prior art memory cell;

FIG. 5 is a figure showing the structure of a six transistor memory cell according to a first embodiment of the present invention;

FIG. 6 is a figure showing the structure of a storage device in which a number of these memory cells are arranged in an array;

FIG. 7 is a figure showing the structure of a storage transistor of this memory cell, and also showing its operation and the condition of voltage application thereto during writing;

FIG. 8 is a figure showing the condition of voltage application to this memory cell when writing data to its storage transistors;

FIG. 9 is a figure showing the data erase voltage application condition for this storage transistor;

FIG. 10 is a figure showing the condition of voltage application to this memory cell when erasing data from its storage transistors;

FIGS. 11A to 11C are figures for explanation of the data setting method and the reading out margin for this memory cell;

FIG. 12 is a figure showing the condition of voltage application to a storage transistor of this memory cell when reading out data therefrom;

FIG. 13 is a figure showing the condition of voltage application to this memory cell when reading out data from its storage transistors;

FIG. 14 is a structural diagram of a memory cell in which the channel width of one of its load transistors is made wider, so that the continuity resistances of the two inverters are unbalanced;

FIG. 15 is a structural diagram of a memory cell in which extra condensers are connected, so that the electrostatic capacities possessed by the two inverters are unbalanced;

FIG. 16 is a figure showing the structure of a VPS divided type eight transistor memory cell, according to a second embodiment of the present invention;

FIG. 17 is a figure showing the structure of a storage device in which a number of these memory cells are arranged in an array;

FIG. 18 is a figure showing the condition of voltage application when writing data to a storage transistor of this memory cell;

FIG. 19 is a figure showing the condition of voltage application to this memory cell when writing data to its storage transistors;

FIG. 20 is a figure showing the data erase voltage application condition for this storage transistor;

FIG. 21 is a figure showing the condition of voltage application to this memory cell when writing data to its storage transistors;

FIGS. 22A to 22C are figures for explanation of the data setting method and the reading out margin for this memory cell;

FIG. 23 is a figure for explanation of a data determination procedure for this memory cell (which determines the data to “1” even if writing has not yet been performed);

FIGS. 24A to 24C are figures for explanation of the reading out margin for data in the data determination procedure described above;

FIG. 25 is a figure for explanation of a procedure for determining the data of this memory cell, which is in its initial state, to “1”;

FIG. 26 is a figure showing the condition of voltage application when reading out data from a storage transistor of this memory cell;

FIG. 27 is a figure showing the condition of voltage application when reading out data from the storage transistors of this memory cell;

FIG. 28 is a figure for explanation of a data determination procedure for this memory cell;

FIG. 29 is a figure for explanation of another data determination procedure for this memory cell;

FIG. 30 is a figure for explanation of yet another data determination procedure for this memory cell;

FIG. 31 is a structural diagram of a memory cell in which the channel width of one of its load transistors is made wider, so that the continuity resistances of the two inverters are unbalanced;

FIG. 32 is a structural diagram of a memory cell in which extra condensers are connected, so that the electrostatic capacities possessed by the two inverters are unbalanced;

FIG. 33 is a figure for explanation of a method of measuring the threshold voltages of the storage transistors;

FIG. 34 is a figure showing the structure of a source line segmented type eight transistor memory cell, according to a third embodiment of the present invention;

FIG. 35 is a figure showing the structure of a storage device in which a number of these memory cells are arranged in an array;

FIG. 36 is a figure showing the condition of voltage application when writing data to a storage transistor of this memory cell;

FIG. 37 is a figure showing the condition of voltage application to this memory cell when writing data to its storage transistors;

FIG. 38 is a figure showing the data erase voltage application condition for this storage transistor;

FIG. 39 is a figure showing the condition of voltage application to this memory cell when writing data to its storage transistors;

FIGS. 40A to 40C are figures for explanation of the data setting method and the reading out margin for this memory cell;

FIG. 41 is a figure for explanation of a data determination procedure for this memory cell (which determines the data to “1” even if writing has not yet been performed);

FIGS. 42A to 42C are figures for explanation of the reading out margin for data in the data determination procedure described above;

FIG. 43 is a figure for explanation of a procedure for determining the data of this memory cell, which is in its initial state, to “1”;

FIG. 44 is a figure showing the condition of voltage application when reading out data from a storage transistor of this memory cell;

FIG. 45 is a figure showing the condition of voltage application when reading out data from the storage transistors of this memory cell;

FIG. 46 is a figure for explanation of a data determination procedure for this memory cell;

FIG. 47 is a figure for explanation of another data determination procedure for this memory cell;

FIG. 48 is a figure for explanation of yet another data determination procedure for this memory cell;

FIG. 49 is a structural diagram of a memory cell in which the channel width of one of its load transistors is made wider, so that the continuity resistances of the two inverters are unbalanced;

FIG. 50 is a structural diagram of a memory cell in which extra condensers are connected, so that the electrostatic capacities possessed by the two inverters are unbalanced;

FIG. 51 is a figure for explanation of a method of measuring the threshold voltages of the storage transistors;

FIG. 52 is a figure showing the layout upon the semiconductor substrate of the memory cells according to the second and the third embodiments;

FIG. 53 is a structural diagram of an RFID chip which employs the memory cell described above; and

FIG. 54 is a schematic plan view of a system LSI chip which employs the memory cell described above.

DETAILED DESCRIPTION OF THE INVENTION

Embodiment One

A non-volatile storage device according to a first embodiment of the present invention, and a semiconductor integrated circuit device which incorporates it, will now be explained with reference to FIGS. 5 through 15.
FIG. 5 is a circuit diagram of a single non-volatile memory cell of a non-volatile storage device. This memory cell comprises six MOS transistors. It includes a flip-flop in which an inverter (a true side inverter) in which a P-type MOS transistor MP1 and an N-type MOS transistor MCN1 are connected in series, and an inverter (a bar side inverter) in which a P-type MOS transistor MP2 and an N-type MOS transistor MCN2 are connected in series, are static latch connected. Among these, the P-type MOS transistors MP1 and MP2 are termed load transistors, while the N-type MOS transistors MCN1 and MCN2 are termed storage transistors. As explained in FIG. 7 below, the storage transistors MCN1 and MCN2 function as non-volatile elements which can change their threshold values in a non-volatile manner by accumulating and neutralizing electric charges on their side spacer portions.
In this flip-flop, the inverter in which the load transistor MP1 and the storage transistor MCN1 are connected in series functions as a storage unit on the true side, while the inverter in which the load transistor MP2 and the storage transistor MCN2 are connected in series functions as a storage unit on the bar side. The connection portion between the load transistor MP1 and the storage transistor MCN1 is a node T, while the connection portion between the load transistor MP2 and the storage transistor MCN2 is a node B. When the node T is at high potential and the node B is at low potential, the stored contents is “0”; while, when the node T is at low potential and the node B is at high potential, the stored contents is “1”.
The storage transistor side end portions of these inverters, in other words the sources of the storage transistors MCN1 and MCN2, are connected to a source line SL (a supply line for a first voltage). And the load transistor side end portions of these inverters, in other words the sources of the load transistors MP1 and MP2, are connected to a VPS line (a supply line for a second voltage). Furthermore, the wells of the load transistors MP1 and MP2 are connected to a VPM line.
The node T is connected to a bit line BLT (Bit Line-True) via a transfer gate MN1, while the node B is connected to a bit line BLB (Bit Line-Bar) via a transfer gate MN2. The transfer gates MN1 and MN2 are N-type MOS transistors, and a common word line WL is connected to both of these gates.
FIG. 6 is a figure showing a structure in which a number of the non-volatile cells shown in FIG. 5 are arranged in an array. In this memory cell array, the non-volatile memory cells of FIG. 5 are arranged in an X,Y matrix array. And one word line WL is provided for each row (the rows are arranged along the Y axis direction), while one bit line BLT and one bit line BLB are provided for each column (the columns are arranged along the X axis direction). Each of these word lines WL and bit lines BLT and BLB is controlled independently. On the other hand, the other signal lines (VPS, VPM, and SL) are provided in common for all of the memory cells, and are controlled all together for the entire memory cell array.
FIG. 7 shows the structure of one of these storage transistors MCN1 (or MCN2), and its operation during writing. Furthermore, FIG. 8 is a figure showing the state of voltage application when writing to this storage transistor MCN1.
In FIG. 7, a P-type well 104 of depth 0.8 μm and average boron density 2·10¹⁷cm⁻³is formed upon a surface region of a P-type silicon substrate 101 of resistivity 10 Ω·cm. The two separate storage transistors MCN1 and MCN2 are defined by a plurality of trenches (element separators) 102 of depth 250 nm on this P-type well 104. In this figure, only one of the storage transistors (MCN1) is shown.
This storage transistor is an N-channel type transistor, and comprises, on the surface region of the P-type well 104, a drain 109 and a source 115 which are formed adjacent to the trenches 102 on the two sides, and a drain extension 107 which is formed in a region adjacent to the drain 109. The drain 109 and the source 115 are formed with an average arsenic density of 1·10²⁰cm⁻³, and the drain extension 107 is formed with an average arsenic density of 5·10¹⁸cm⁻³.
Furthermore, upon the substrate in the channel region, which is the region of the surface of the P-type well 104 between the drain 109 and the source 115, there are formed a gate oxide layer 105 of thickness 5 nm, and a gate electrode 106 which is made from a polysilicon layer of thickness 200 nm and having a phosphorus density of 2·10²⁰cm⁻³. Furthermore, on both sides of this gate oxide layer 105 and gate electrode 106, there are formed side spacers 108 which are made as insulating layers of thickness 50 nm. It should be understood that the side spacer 108S on the source side is exposed to the channel region of the substrate, since no extension region is formed around the periphery of the source 115.
Furthermore, within the region of the P-type well 104, a P-type diffusion layer 111 having an average boron density of 1·10²⁰cm⁻³, which is an electrode for grounding this P-type well, is formed in a region which is separated from the storage transistor described above by one of the trenches 102.
With this storage transistor, the threshold voltage can be elevated by injecting carriers into the side spacer 108S on the source side. Furthermore, as will be explained with reference to FIG. 9, it is also possible to return the threshold voltage to its initial state by extracting the carriers which have been injected into the side spacer 108S. In this way, this storage transistor stores data in a non-volatile manner.
It should be understood that, although the initial threshold voltage of this storage transistor is 1.2 V, the variation is great since it is a transistor of a distinctive structure, and accordingly, from the point of view of reliability, it is not possible to utilize such a storage transistor singly as a storage element. Because of this, in this embodiment, the memory cell is built with the flip-flop structure shown in FIG. 5.
In FIG. 7, the write operation is an operation of injecting channel hot electrons HE into the aforementioned side spacer 108S by applying 0 V to a drain line VD and by applying to a source line VS a positive voltage (for example 6V) which is less than the junction withstand voltage. Due to this injection of channel hot electrons, the threshold voltage is elevated by trapped electrons, and the system transmits to the written state.
In order to bring the storage transistor MCN1 to this potential configuration, voltages are applied to the memory cell in the condition shown in FIG. 8. This figure shows the state when the threshold voltage of the storage transistor MCN1 is elevated and the data “0” is being written. In the condition with VPS, VPM, and SL set to 6 V, the word line WL is set to Vcc (1.8 V), and the bit line BLT is set to 0 V while BLB is set to Vcc. By putting ON the transfer gate MN1, which is a N-type MOS transistor, along with the node T (the drain of the storage transistor MCN1) being brought to almost the same potential as the bit line BLT (0 V), since, due to this, the load transistor MP2 is turned ON, accordingly the node B (the gate of the storage transistor MCN1) is brought to almost the same potential as VPS (6 V). Due to this, the voltage application condition of the storage transistor MCN1 becomes the same as in FIG. 7. At this time, electrical currents of around 300 μA flow in the transfer gate MN1 and the storage transistor MCN1, and the threshold voltage of the storage transistor MCN1 is elevated to Vth2.
Furthermore, when writing the data “1”, while the threshold voltage on the side of the storage transistor MCN2 becomes elevated, the other conditions are the same as when writing the data “0”, with only the voltage settings for BLT=Vcc and BLB=0 V being reversed.
It should be understood that although, in the embodiment described above, 6V was applied to the gate of the transistor MCN1 (the node B), and 6V was also applied to the source of the transistor MCN1 (the source line SL), it would also be acceptable for these voltages to be different voltages.
FIGS. 9 and 10 are figures showing the conditions in which erase voltages are applied to the storage transistors MCN1 (MCN2). Referring to FIG. 9, the erase operation is an operation in which a positive voltage less than the junction withstand voltage (for example 9 V) is applied to the source line VS, and 0 V is applied to the gate line VG and to the drain line VD, and thereby an avalanche of hot holes HH is created from the source electrode 115, and these are injected into the side spacer 108S. Due to this, the electrons in the side spacer 108S which were trapped by the above described write operation are neutralized, and the data which was written is erased by decreasing the threshold voltage.
In the actual memory cell, these voltages are applied under the conditions shown in FIG. 10. Fundamentally, the erase operation is performed all at once for all the memory cells in the memory array shown in FIG. 6. Under the conditions that VPM is set to Vcc, VPS is set to 0 V, and SL is set to 9 V, the word lines WL are brought to Vcc and the bit lines BLT and BLB are brought to 0 V. By turning ON the transfer gates MN1 and MN2, which are N-type MOS transistors, the node T and the node B are brought to 0 V, and the voltage application conditions of the storage transistors MCN1 and MCN2 are brought to be the same as those shown in FIG. 9.
Since, in this manner, with the voltage application conditions during writing as shown in FIGS. 7 and 8, and with the voltage application conditions during erasure as shown in FIGS. 9 and 10, the design is arranged so that these operations are performed with the word lines and the bit lines, for which independent control is required for each memory cell, being brought to 0 V or to Vcc, in other words so that the application of high voltage is unnecessary, accordingly it becomes unnecessary to use transistors of high withstand voltage in the control circuitry for the word lines and the bit lines, and thereby it is possible to anticipate the use of transistors of higher performance than in the case of the prior art, whereby the speed of the reading out operation may be increased.
FIGS. 11A to 11C are figures for explanation of the method of threshold value control for the storage transistors MCN1 and MCN2 by the above described write operation, in other words of the method of setting data into this non-volatile memory cell. Here, the data is “1” when the threshold voltage of the storage transistor MCN1 is in the low state (ON) and moreover the threshold voltage of the storage transistor MCN2 is in the high state (OFF); while the data is “0” when the threshold voltage of the storage transistor MCN1 is in the high state (OFF) and moreover the threshold voltage of the storage transistor MCN2 is in the low state (ON).
FIG. 11A shows the case before data setting, in other words when the threshold voltages of both of the storage transistors MCN2 and MCN2 are in their initial state Vth0. In this case, the state of the memory cell is indeterminate.
FIG. 11B shows the threshold voltages when this non-volatile memory cell is set to the data “0”. The writing of the data “0” is implemented by raising the threshold voltage of the storage transistor MCN1 from its initial state of FIG. 11A to Vth2 (where Vth2>Vth0).
And FIG. 11 c shows the threshold voltages when this non-volatile memory cell is set to the data “1”. The writing of the data “1” is implemented by raising the threshold voltage of the storage transistor MCN2 from its initial state of FIG. 11A to Vth2 (where Vth2>Vth0).
When the erase operation explained with reference to FIGS. 9 and 10 is performed, the system returns to the state shown in FIG. 11A of this figure, even though the threshold voltages are controlled as shown in FIGS. 11B and 11C.
Since, in this manner, the structure is arranged so that, even though the threshold voltages of the storage transistors MCN1 and MCN2 have once been raised, it is still possible again to lower them to the initial state Vth0, accordingly, even though requests for rewriting of the data have been issued a number of times, it is still possible to obtain a sufficient margin for reading out, which is the difference between the threshold voltages on the true side (the storage transistor MCN1) and on the bar side (the storage transistor MCN2).
FIGS. 12 and 13 are figures showing the voltage application conditions when reading out this non-volatile memory cell. Referring to FIG. 12, the operation of reading out is performed by reading out the voltage on the drain line VD, when the source line VS is set to 0 V and Vcc has been applied to the gate line VG. In the actual memory cell, the voltages are applied in the conditions shown in FIG. 13. This read out operation is the same as the operation of reading out a SRAM, and employs a differential type sense amp. VPS and VPM are set to Vcc and SL is set to 0 V, and the change in the bit lines BLT and BLB, which corresponds to the data in this flip-flop, is read out by the differential type sense amp, under the condition that the word line WL is set to Vcc. When BLT is at a low voltage (0 V) and moreover BLB is at a high voltage (Vcc), then the data is “1”; while, when BLT is at a high voltage (Vcc) and moreover BLB is at a low voltage (0 V), then the data is “0”.

Variants of Embodiment One

With this six transistor non-volatile memory cell of the first embodiment, in the state in which writing has not been performed, in other words in the state in which the threshold voltages of the storage transistors MCN1 and MCN2 are both low, the data is indeterminate. Thus, as shown in FIGS. 14 and 15, by unbalancing the circuit structure on the true side (the storage transistor MCN1) and on the bar side (the storage transistor MCN2), it is possible to determine the data as being “0” or “1” when the power is raised.
The non-volatile memory cell shown in FIG. 14 is one in which the channel widths of the two load transistors MP1 and MP2 are unbalanced. In this example, the channel width of the load transistor MP1 (shown by the thick lines) is made to be twice the channel width of the load transistor MP2, so that the resistance value of the load transistor MP1 when it is ON becomes about ½ of the resistance value of the load transistor MP2 when it is ON.
When the power supply to the memory cell having this structure is turned ON in the state in which neither one of the storage transistors MCN1 and MCN2 has been written, the potential at the node T rises faster than that at the node B, and the load transistor MP1 and the storage transistor MCN2 go into the ON state, while the load transistor MP2 and the storage transistor MCN1 go into the OFF state; in other words, the data stabilizes at “0”.
It should be understood that, instead of unbalancing the channel widths as described above, it would also be acceptable to unbalance the channel lengths. Furthermore, it would also be acceptable for the load transistor whose channel width or channel length is changed to be either MP1 or MP2. Moreover, it would also be acceptable to perform this unbalancing by changing the channel width or the channel length of one of the storage transistors MCN1 or MCN2.
On the other hand, the non-volatile memory cell shown in FIG. 15 is one in which a capacitor is connected to each of the two nodes T and B of the flip-flop. A capacitor C1 is connected between the node T and the power supply line Vcc, while a capacitor C2 is connected between the node B and ground. The capacitances of these capacitors may be, for example, around 50 fF.
Due to this, when the power supply is turned ON in the state in which neither writing nor erasure of either of the two storage transistors MCN1 and MCN2 has been performed, since the potential of the node T rises quickly directly after the power supply has been turned ON while the potential of the node B rises more slowly, accordingly the system stabilizes with the load transistor MP1 and the storage transistor MCN2 in the ON state and the load transistor MP2 and the storage transistor MCN1 in the OFF state, in other words at the data “0”.
Although, in the example shown in FIG. 15, capacitors were connected to both the node T and to the node B, it would also be possible to provide a non-volatile memory cell of an asymmetric circuit structure, with a capacitor being connected to only one of the nodes.

Embodiment Two

A non-volatile storage device according to a second embodiment of the present invention, and a semiconductor integrated circuit device which incorporates it, will now be explained with reference to FIGS. 16 through 33.
FIG. 16 is a circuit diagram of a single non-volatile memory cell of this non-volatile storage device. This memory cell is a non-volatile memory cell of a VPS divided eight transistor type. This memory cell comprises a flip-flop in which an inverter (a true side inverter) in which a P-type MOS transistor MP1 and an N-type MOS transistor MCN1 are connected in series, and an inverter (a bar side inverter) in which a P-type MOS transistor MP2 and an N-type MOS transistor MCN2 are connected in series, are static latch connected. Among these, the P-type MOS transistors MP1 and MP2 are termed load transistors, while the N-type MOS transistors MCN1 and MCN2 are termed storage transistors. As explained in FIG. 18 below, the storage transistors MCN1 and MCN2 function as non-volatile elements which can change their threshold values in a non-volatile manner by accumulating and neutralizing electric charges on their side spacer portions.
The storage transistor side end portions of these inverters, in other words the sources of the storage transistors MCN1 and MCN2, are connected to a source line SL. And the load transistor side end portion of the true side inverter, in other words the source of the load transistor MP1, is connected to a line VPST, while the load transistor side end portion of the bar side inverter, in other words the source of the load transistor MP2, is connected to a line VPSB.
In this flip-flop, the inverter in which the load transistor MP1 and the storage transistor MCN1 are connected in series functions as a storage unit on the true side, while the inverter in which the load transistor MP2 and the storage transistor MCN2 are connected in series functions as a storage unit on the bar side. The connection portion between the load transistor MP1 and the storage transistor MCN1 is a node T, while the connection portion between the load transistor MP2 and the storage transistor MCN2 is a node B. When the node T is at high potential and the node B is at low potential, the stored contents is “0”; while, when the node T is at low potential and the node B is at high potential, the stored contents is “1”.
The node T is connected to a bit line BLT (Bit Line-True) via a transfer gate MN1, while the node B is connected to a bit line BLB (Bit Line-Bar) via a transfer gate MN2. The transfer gates MN1 and MN2 are N-type MOS transistors, and a common word line WL is connected to both of these gates.
Furthermore a P-type MOS transistor MP3, which is a transistor for pre-charge, is connected in parallel with the load transistor MP1, in other words between the node T and VPST. Moreover a P-type MOS transistor MP4, which also is a transistor for pre-charge, is connected in parallel with the load transistor MP2, in other words between the node B and VPSB. A T-side pre-charge line PRET is connected to the gate of the P-type MOS transistor MP3, while a B-side pre-charge line PREB is connected to the gate of the P-type MOS transistor MP4. Furthermore, all of the P-type MOS transistors MP1 through MP4 are formed within the same N well, and the potential of this N well is controlled by a signal VPM.
FIG. 17 is a figure showing a structure in which a number of the non-volatile cells shown in FIG. 16 are arranged in an array. In this memory cell array, the non-volatile memory cells of FIG. 16 are arranged in an X,Y matrix array. And one word line WL is provided for each row (the rows are arranged along the Y axis direction), while one bit line BLT and one bit line BLB are provided for each column (the columns are arranged along the X axis direction). Each of these word lines WL and bit lines BLT and BLB is controlled independently. On the other hand, the other signal lines (PREB, PRET, VPST, VPSB, VPM, and SL) are provided in common for all of the memory cells, and are controlled all together for the entire memory cell array.
FIG. 18 shows the structure of one of these storage transistors MCN1 (or MCN2), and its operation during writing. Furthermore, FIG. 19 is a figure showing the state of voltage application when writing to this storage transistor MCN1. In FIG. 18, a P-type well 104 of depth 0.8 μm and average boron density 2·10¹⁷cm⁻³is formed upon a surface region of a P-type silicon substrate 101 of resistivity 10 Ω·cm. The two separate storage transistors MCN1 and MCN2 are defined by a plurality of trenches (element separators) 102 of depth 250 nm on this P-type well 104. In this figure, only one of the storage transistors (MCN1) is shown.
This storage transistor is an N-channel type transistor, and comprises, on the surface region of the P-type well 104, a drain 109 and a source 115 which are formed adjacent to the trenches 102 on the two sides, and a drain extension 107 which is formed in a region adjacent to the drain 109. The drain 109 and the source 115 are formed with an average arsenic density of 1·10²⁰cm³, and the drain extension 107 is formed with an average arsenic density of 5·10¹⁸cm⁻³.
Furthermore, upon the substrate in the channel region, which is the region of the surface of the P-type well 104 between the drain 109 and the source 115, there are formed a gate oxide layer 105 of thickness 5 nm, and a gate electrode 106 which is made from a polysilicon layer of thickness 200 nm and having a phosphorus density of 2·10²⁰cm⁻³. Furthermore, on both sides of this gate oxide layer 105 and gate electrode 106, there are formed side spacers 108 which are made as insulating layers of thickness 50 nm. It should be understood that the side spacer 108S on the source side is exposed to the channel region of the substrate, since no extension region is formed around the periphery of the source 115.
Furthermore, within the region of the P-type well 104, a P-type diffusion layer 111 having an average boron density of 1·10²⁰cm⁻³, which is an electrode for grounding this P-type well, is formed in a region which is separated from the storage transistor described above by one of the trenches 102.
With this storage transistor, the threshold voltage can be elevated by injecting carriers into the side spacer 108S on the source side. Furthermore, as will be explained with reference to FIG. 20, it is also possible to return the threshold voltage to its initial state by extracting the carriers which have been injected into the side spacer 108S. In this way, this storage transistor stores data in a non-volatile manner.
It should be understood that, although the initial threshold voltage of this storage transistor is 1.2 V, the variation is great since it is a transistor of a distinctive structure, and accordingly, from the point of view of reliability, it is not possible to utilize such a storage transistor singly as a storage element. Because of this, in this embodiment, the memory cell is built with the flip-flop structure shown in FIG. 16.
In FIG. 18, the write operation is an operation of injecting channel hot electrons HE into the aforementioned side spacer 108S by applying 0 V to the drain line VD and by applying a positive voltage (for example 6 V) which is less than the junction withstand voltage to the source line VS. Due to this injection of channel hot electrons, the threshold voltage is elevated by trapped electrons, and the system transmits to the write state.
In order to bring the storage transistor MCN1 to this potential configuration, voltages are applied to the memory cell in the condition shown in FIG. 19. This figure shows the state when the threshold voltage of the storage transistor MCN1 is elevated and the data “0” is being written. In the condition with PREB, PRET, VPST, VPSB, VPM, and SL set to 6 V, the word line WL is set to Vcc, and the bit line BLT is set to 0 V while BLB is set to Vcc. By putting ON the transfer gate MN1, which is a N-type MOS transistor, along with the node T (the drain of the storage transistor MCN1) being brought to almost the same potential as the bit line BLT (0 V), since, due to this, the load transistor MP2 is turned ON, accordingly the node B (the gate of the storage transistor MCN1) is brought to almost the same potential as VPSB (6 V). Due to this, the voltage application condition of the storage transistor MCN1 becomes the same as in FIG. 18. At this time, electrical currents of around 300 μA flow in the transfer gate MN1 and in the storage transistor MCN1, and the threshold voltage of the storage transistor MCN1 is elevated to Vth2.
Furthermore, when writing the data “1”, while the threshold voltage on the side of the storage transistor MCN2 becomes elevated, the other conditions are the same as when writing the data “0”, with only the voltage settings for BLT=Vcc and BLB=0 V being reversed.
It should be understood that although, in the embodiment described above, 6V was applied to the gate of the transistor MCN1 (the node B), and 6V was also applied to the source of the transistor MCN1 (the source line SL), it would also be acceptable for these voltages to be different voltages.
FIGS. 20 and 21 are figures showing the conditions in which erase voltages are applied to the storage transistors MCN1 (MCN2). Referring to FIG. 20, the erase operation is an operation in which a positive voltage less than the junction withstand voltage (for example 9 V) is applied to the source line VS, and 0 V is applied to the gate line VG and to the drain line VD, and thereby an avalanche of hot holes HH from the source electrode 115 is created, and these are injected into the side spacer 108S. Due to this, the electrons in the side spacer 108S trapped by the above described write operation are neutralized, and the data which was written is erased by decreasing the threshold voltage.
In the actual memory cell, these voltages are applied under the conditions shown in FIG. 21. Fundamentally, the erase operation is performed all at once for all the memory cells in the memory array shown in FIG. 17. Under the conditions that PREB, PRET, and VPM are set to Vcc, VPST and VPSB are set to 0 V, and SL is set to 9 V, the word lines WL are brought to Vcc and the bit lines BLT and BLB are brought to 0 V. By turning ON the transfer gates MN1 and MN2, which are N-type MOS transistors, the node T and the node B are brought to 0 V, and the voltage application conditions of the storage transistors MCN1 and MCN2 are brought to be the same as those shown in FIG. 20.
Since, in this manner, with the voltage application conditions during writing as shown in FIGS. 18 and 19, and with the voltage application conditions during erasure as shown in FIGS. 20 and 21, the design is arranged so that these operations are performed with the word lines and the bit lines, for which independent control is required for each memory cell, being brought to 0 V or to Vcc, in other words so that the application of high voltage is unnecessary, accordingly it becomes unnecessary to use transistors of high withstand voltage in the control circuitry for the word lines and the bit lines, and thereby it is possible to anticipate the use of transistors of higher performance than in the prior art, whereby the speed of the reading out operation may be increased.
FIGS. 22A to 22C are figures for explanation of the method of threshold value control for the storage transistors MCN1 and MCN2 by the above described write operation, in other words of the method of setting data into this non-volatile memory cell. Here, the data is “1” when the threshold voltage of the storage transistor MCN1 is in the low state (ON) and moreover the threshold voltage of the storage transistor MCN2 is in the high state (OFF); while the data is “0” when the threshold voltage of the storage transistor MCN1 is in the high state (OFF) and moreover the threshold voltage of the storage transistor MCN2 is in the low state (ON).
FIG. 22A shows the case before data setting, in other words when the threshold voltages of both of the storage transistors MCN1 and MCN2 are in their initial state Vth0. Even in this case, due to the procedure shown in FIG. 23 or FIG. 25, the state of this non-volatile memory cell is determined as being data “1”.
FIG. 22B shows the threshold voltages when this non-volatile memory cell is set to the data “0”. The writing of the data “0” is implemented by raising the threshold voltage of the storage transistor MCN1 from its initial state of (A) of this figure to Vth2 (where Vth2>Vth0).
And FIG. 22C shows the threshold voltages when this non-volatile memory cell is set to the data “1”. The writing of the data “1” is implemented by raising the threshold voltage of the storage transistor MCN2 from its initial state of (A) of this figure to Vth2 (where Vth2>Vth0).
When the erase operation explained with reference to FIGS. 20 and 21 is performed, the system returns to the state shown in FIG. 22A of this figure, even though the threshold voltages are controlled as shown in FIGS. 22B and 22C.
Since, in this manner, the structure is arranged so that, even though the threshold voltages of the storage transistors MCN1 and MCN2 have once been raised, it is still possible again to lower them to the initial state Vth0, and furthermore since it is possible forcibly to determine the storage transistors MCN1 and MCN2 to the data “1”, even if they are both in the initial state Vth0, accordingly, even though requests for rewriting of the data have been issued a number of times, it is still possible to obtain a sufficient margin for reading out, which is the difference between the threshold voltages on the true side (the storage transistor MCN1) and on the bar side (the storage transistor MCN2).
By the writing operation shown in FIGS. 18 and 19, it is possible to perform writing to either one of the storage transistors MCN1 and MCN2, and it is possible to write the data “1” or “0” into the memory cell. On the other hand, if the memory cell is in its initial state, in other words if writing has not been performed to either one of the storage transistors MCN1 and MCN2 so that the threshold voltages of both of the storage transistors MCN1 and MCN2 are in their initial state Vth0, then generally the contents stored in the memory cell is indeterminate. However, by applying voltages to the memory cell in its initial state according to the following procedure, it is possible to determine the storage contents of this memory cell to either “1” or “0”.
FIG. 23 is a figure for explanation of the voltage application procedure for determining the data of the non-volatile memory cell of the present invention. This procedure is an operation which, if the storage contents of this non-volatile memory cell is “1” or “0”, can set the state of the memory cell (the flip-flop) according to this storage contents, and which can fix the data forcibly at “1” if the memory cell is in its initial state. If, in a memory array in which a plurality of memory cells are arranged, memory cells whose storage contents is “1” and “0” and memory cells in the initial state are mixed together, then, by performing this procedure all at once for all of the memory cells in the memory array, for those memory cells whose storage contents is “1” and “0”, the states of their flip-flops are set according to their storage contents, while, for those memory cells which are in the initial state, the states of their flip-flops are forcibly set to “1”. This procedure is executed during starting up of the memory.
The procedure shown in FIG. 23 proceeds as follows. In the condition with the word line WL and the bit lines BLT and BLB set to 0 V, first, the source potential SL is raised from 0 V to Vcc at the time instant T0, so that the storage transistors MCN1 and MCN2 are both turned OFF; and then, at the time instant T1, the pre-charge transistors MP3 and MP4 are turned ON by dropping the pre-charge control signals PRET and PREB from Vcc to 0 V. At the same time, among the pre-charge voltages which are set to Vcc, the pre-charge voltage VPST on the true side is decreased by ΔV (which is a voltage sufficiently smaller than Vth2−Vth0, for example 0.2 V). Due to this, the pre-charge voltages at the node T and the node B also become Vcc−ΔV and Vcc respectively, and the voltage between the source and the drain of the storage transistor MCN2 becomes just ΔV higher than the voltage between the source and the drain of the storage transistor MCN1. Because of this, it is possible to increase the apparent threshold voltage on the side of the storage transistor MCN2 by just ΔV.
At the time instant T2, PRET and PREB return to Vcc, and, from the time instant T3, the source potential SL slowly drops towards 0 V. At this time, if the storage contents of the memory cell is “1”, then, since the transistor MCN1 whose threshold voltage is the lower goes to ON before the transistor MCN2, accordingly the node T is pulled down to 0 V first while the node B goes to Vcc, so that the data of the flip-flop is determined at “1”. On the other hand, if the storage contents of the memory cell is “0”, then, since likewise the threshold voltage of the transistor MCN2 is lower than the threshold voltage of the transistor MCN1 (even though its threshold voltage is elevated by just ΔV) so that the storage transistor MCN2 goes to ON first, accordingly the node B is pulled down to 0 V first while the node T goes to Vcc, so that the data of the flip-flop is determined at “0”.
Furthermore, if the storage contents of the memory cell is “indeterminate”, in other words if the threshold voltages of both the transistors MCN1 and MCN2 are Vth0, then, since the apparent threshold voltage of the storage transistor MCN1 is just ΔV lower than the threshold voltage of the storage transistor MCN2, accordingly the transistor MCN1 goes ON before the transistor MCN2, and the node T is pulled down to 0 V first while the node B goes to Vcc, so that the data of the flip-flop is determined at “1”.
VPST is returned to Vcc at a time instant T4 after the state of the flip-flop is determined.
Since often a memory cell in which the threshold voltages of MCN1 and MCN2 are both Vth0 is a cell to which, up till now, neither writing nor rewriting has been performed, and since, with this type of memory cell, the transistors do not deteriorate along with rewriting, accordingly, with regard to the setting of ΔV, it is sufficient only to pay consideration to variation of the initial threshold voltages of the transistors. Thus it is considered that, for example, around 0.2 V is sufficient.
The voltage application procedure shown in FIG. 23 is the procedure when determining the memory cell in its initial state to the data “1”, but it is also possible to determine the memory cell in its initial state to the data “0” by, in the source potential control of the P-type MOS transistors, decreasing the voltage of VPSB by just ΔV, instead of decreasing the voltage of VPST as above.
FIGS. 24A to 24C are figures for explanation of the margin for data determination when performing the voltage application procedure shown in FIG. 23. In the initial state in which the threshold voltages of the transistors MCN1 and MCN2 are both Vth0, as previously described, by making the pre-charge voltage of the node T be just ΔV lower than that of the node B, the apparent threshold voltage on the MCN2 side is increased by just ΔV, so that the data is forcibly recognized as “1” (FIG. 24A). Although, for a memory cell in which the data “0” has already been written, the margin comes to be decreased by just ΔV, suppose that Vth2−Vth0=1 V and ΔV=0.2 V: then the margin becomes 0.8 V (FIG. 24B). And conversely, for a memory cell in which the data “1” has already been written, by increasing the margin by just ΔV, suppose that Vth2−Vth0=1 V and ΔV=0.2 V: then the margin becomes 1.2 V (FIG. 24C).
FIG. 25 is a figure for explanation of the procedure for voltage application for determining the data for a memory cell which is in its initial state. In other words, while the procedure shown in FIG. 23 is a procedure which, for a memory array in which memory cells whose storage contents are “1” and “0” and memory cells which are in their initial state are mixed together, determines the data of these memory cells all at once, by contrast, the procedure shown in FIG. 25 is a procedure which is only effective for a memory cell which is in its initial state, and which reliably determines the data of such a memory cell in its initial state as being either “1” or “0”.
By executing this procedure, if memory cells to which data has already been written and memory cells which are in their initial state are mixed together, individually for those memory cells which are in their initial states, or by executing this procedure all at once for a memory array of which all of the memory cells are in their initial state, it is possible to determine the data of those memory cells as “1” or “0”. In the following explanation, the procedure will be shown for forcibly determining the storage contents of a memory cell in its initial state to “1”.
Under the condition that the word line WL and the bit lines BLT and BLB are set to 0 V, first at the time instant T0 the source potential SL is raised from 0 V to Vcc and MCN1 and MCN2 are turned OFF, and then at the time instant T1, by dropping the pre-charge control signals PREB and PRET and the source potential VPST of the pre-charge transistor MP3 from Vcc to 0 V, the node B is charged up to Vcc by the pre-charge transistor MP4, and the node T is electrically discharged via the pre-charge transistor MP3. And, at the time instant T2, PREB and PRET return to Vcc, and then, at the time instant T3, by setting SL from Vcc to 0 V, due to the transistor MCN1 whose gate potential is the higher going to ON before the transistor MCN2, the node T is pulled down to 0 V in the state in which the node B is held at Vcc, so that the data in this flip-flop is determined as being “1”. And then, at the time instant T4, VPST is returned to Vcc.
The procedure shown in FIG. 25 is a procedure for forcibly determining the storage contents of the memory cell to “1”, but it would also be possible forcibly to determine the storage contents to “0”. In other words, by dropping the source potential VPSB of the pre-charge transistor MP4 from Vcc to 0 V at the time instant T1, instead of the source potential VPST of the pre-charge transistor MP3, the node B is set to 0 V, so that the data is determined at “0”.
By executing the procedure of FIG. 23 or the procedure of FIG. 25, even if memory cells are present in their initial state in which both of the two storage transistors MCN1 and MCN2 are in their initial state of Vth0, it is possible to determine their storage contents uniquely at “1” or at “0”.
If this non-volatile memory cell is used instead of a fuse, and if its storage contents are uniquely determined to “1”, then, still in the initial state, it is possible to implement the state before cutting the fuse (corresponding to the data “1”).
FIGS. 26 and 27 are figures showing the voltage application conditions when reading out from this non-volatile memory cell. Referring to FIG. 26, the operation of reading out is performed by reading out the voltage on the drain line VD, when the source line VS is set to 0 V and Vcc has been applied to the gate line VG. In the actual memory cell, the voltages are applied in the conditions shown in FIG. 27. This read out operation is the same as the operation of reading out an SRAM, and employs a differential type sense amp. PREB, PRET, VPST, VPSB, and VPM are set to Vcc and SL is set to 0 V, and the change in the bit lines BLT and BLB, which corresponds to the data in this flip-flop, is read out by the differential type sense amp, under the condition that the word line WL is set to Vcc. When BLT is at a low voltage (0 V) and moreover BLB is at a high voltage (Vcc), then the data is “1”; while, when BLT is at a high voltage (Vcc) and moreover BLB is at a low voltage (0 V), then the data is “0”.
While the data determination procedure shown in FIG. 25 shows a method which, even if memory cells in their initial state (i.e. memory cells for which both of the storage transistors MCN1 and MCN2 are in their initial state (with threshold voltage=Vth0)) are present, can forcibly determine them to “1”, if no such memory cells in their initial state are present, or if it will be no problem even if the data values of the memory cells in their initial state are indeterminate, then it is possible to employ the determination procedure shown in FIGS. 28 through 30. With this procedure, no decrease (refer to FIG. 24) of the determination margin by ΔV takes place, as in the procedure shown in FIG. 23.
The procedure shown in FIG. 28 is conducted as follows. Under the condition that the word line WL and the bit lines BLT and BLB are set to 0 V and VPST, VPSB, and VPM are set to Vcc, first at the time instant T0 the source potential SL is raised from 0 V to Vcc, and the storage transistors MCN1 and MCN2 are turned OFF. In this state, by lowering the pre-charge control signals PRET and PREB from Vcc to 0 V at the time instant T1, the pre-charge transistors MP3 and MP4 are turned ON, and the node T and the node B are pre-charged to Vcc. After PRET and PREB return to Vcc at the time instant T2 and the pre-charging has been completed, from the time instant T3, the source potential SL slowly sinks towards 0 V. At this time, if the data is “1”, then, since the transistor MCN1 whose threshold voltage is the lower goes to ON first before the transistor MCN2, accordingly the node T is pulled down to 0 V first while the node B goes to Vcc, so that the data of the flip-flop is determined at “1”. On the other hand, if the data is “0”, then, since the transistor MCN2 whose threshold voltage is the lower goes to ON first before the transistor MCN1, accordingly the node B is pulled down to 0 V first while the node T goes to Vcc, so that the data of the flip-flop is determined at “0”.
And the procedure shown in FIG. 29 is conducted as follows. The word line WL and the bit lines BLT and BLB are set to 0 V, and VPST, VPBSB, and VPM are set to Vcc, and moreover the source potential SL is fixed at 0 V. Under these conditions, by dropping the pre-charge control signals PRET and PREB from Vcc to 0 V at the time instant T0, the pre-charge transistors MP3 and MP4 are turned ON, and the node T and the node B are pre-charged. With regard to the pre-charge level of the node T and the node B, the current flow amount which flows as DC in each of the transistors MP1, MP3, and MCN1, or in each of the transistors MP2, MP4, and MCN2, now becomes constant and stabilizes. At this time, if the data is “1”, then, since the current drive force of the transistor MCN1 whose threshold voltage is the lower is greater than that of MCN2, accordingly the node T goes to a lower potential than the node B, and, after completion of the pre-charge at the time instant T1, the data of the flip-flop is determined at “1”. On the other hand, if the data is “0”, then, since the current drive force of the transistor MCN2 whose threshold voltage is the lower is greater than that of MCN1, accordingly the node B goes to a lower potential than the node T, and, after completion of the pre-charge at the time instant T1, the data of the flip-flop is determined at “0”.
To compare this determination procedure with the determination procedure shown in FIG. 28, although it has a demerit from the point of view of consumption of electrical power since a DC feedthrough current flows in the flip-flop during the pre-charge process, nevertheless it has the merit from the point of view of simplicity and ease of control that it is possible to control the source line potentials SL by fixing them at 0 V.
Moreover, the procedure shown in FIG. 30 is conducted as follows. The distinguishing feature of this data determination procedure is that the pre-charge voltages of the node T and the node B are supplied from the bit line side. Along with fixing PRET and PREB at Vcc, the word line WL and the bit lines BLT and BLB are set to Vcc, and, by setting the potentials VPST and VPSB to 0 V, thus cutting off the transistors MP1 through MP4, the node T and the node B are each pre-charged to Vcc−Vthn via the transfer gates MN1 and MN2 from the bit lines BLT and BLB which have been charged up to Vcc. Here Vthn is the threshold voltage of the transistors MN1 and MN2. At the time instant T0 the word line WL is dropped from Vcc to 0 V, and, with the node T and the node B made floating, the data of the flip-flop is determined by the difference between the amounts of electrical charge discharged from the transistors MCN1 and MCN2. If the data is “1”, then, since the current drive force of the transistor MCN1 whose threshold voltage is the lower is greater than that of MCN2, accordingly the node T goes to a lower potential than the node B, and, after the potentials VPST and VPSB have risen from 0 V to Vcc at the time instant T1, the data of this non-volatile memory cell is determined at “1”. On the other hand, if the data is “0”, then, since the current drive amount of the transistor MCN2 whose threshold voltage is the lower is greater than that of MCN1, accordingly the node B goes to a lower potential than the node T, and, after the potentials VPST and VPSB have risen from 0 V to Vcc at the time instant T1, the data of this non-volatile memory cell is determined at “0”. After this, VPS is raised to Vcc at the time instant T1.

Variants of Embodiment Two

With this eight transistor VPS divided type non-volatile memory cell of the second embodiment, in the state in which writing has not been performed, in other words in the state in which the threshold voltages of the storage transistors MCN1 and MCN2 are both low, the data cannot be determined by performing the raising processing (i.e. the data determination processing) shown in FIGS. 28 through 30. Thus, as shown in FIGS. 31 and 32, by unbalancing the circuit structure on the true side (the storage transistor MCN1) and on the bar side (the storage transistor MCN2), it is possible to determine the data as being one of “0” or “1” upon raising.
The non-volatile memory cell shown in FIG. 31 is one in which the channel widths of the two load transistors MP1 and MP2 are unbalanced. In this example, the channel width of the load transistor MP1 (shown by the thick lines) is made to be twice the channel width of the load transistor MP2, so that the resistance value of the load transistor MP1 when it is ON becomes about ½ of the resistance value of the load transistor MP2 when it is ON.
When the power supply to the memory cell having this structure is turned ON in the state in which neither one of the storage transistors MCN1 and MCN2 has been written, the potential at the node T rises faster than that at the node B, and the load transistor MP1 and the storage transistor MCN2 go into the ON state, while the load transistor MP2 and the storage transistor MCN1 go into the OFF state; in other words, the data stabilizes at “0”.
It should be understood that, instead of unbalancing the channel widths as described above, it would also be acceptable to unbalance the channel lengths. Furthermore, it would also be acceptable for the load transistor whose channel width or channel length is changed to be either MP1 or MP2. Moreover, it would also be acceptable to perform this unbalancing by changing the channel width or the channel length of one of the storage transistors MCN1 or MCN2.
On the other hand, the non-volatile memory cell shown in FIG. 32 is one in which a capacitor is connected to each of the two nodes T and B of the flip-flop. A capacitor C1 is connected between the node T and the power supply line Vcc, while a capacitor C2 is connected between the node B and ground. The capacitances of these capacitors may be, for example, around 50 fF.
Due to this, when the power supply is turned ON in the state in which neither writing nor erasure of either of the two storage transistors MCN1 and MCN2 has been performed, since the potential of the node T rises quickly directly after the power supply has been turned ON while the potential of the node B rises more slowly, accordingly the system stabilizes with the load transistor MP1 and the storage transistor MCN2 in the ON state and the load transistor MP2 and the storage transistor MCN1 in the OFF state, in other words at the data “0”.
Although, in the example shown in FIG. 32, capacitors were connected to both the node T and to the node B, it would also be possible to provide a non-volatile memory cell of an asymmetric circuit structure, with a capacitor being connected to only one of the nodes.
—The Threshold Voltage Measurement Method in the Second Embodiment—
It should be understood that, with this non-volatile memory cell of an eight transistor structure, it is possible to measure the threshold voltages of the storage transistors by establishing a potential arrangement as, for example, shown in FIG. 33. By measurement of the threshold voltages by this method, it becomes possible to perform evaluation of the variation of the threshold voltages in the initial state, and evaluation of the amount of change of the threshold voltages during a write operation or an erase operation and of the high temperature storage characteristics of the threshold voltages after rewriting and so on.
In FIG. 33, the case is shown of deciding upon the threshold voltage of the storage transistor MCN1. In the state in which the source potential SL of the storage transistor MCN1=0 V, 0.5 V is supplied from the bit line BLT via the transfer gate MN1 to the drain of this storage transistor MCN1 (i.e. to the node T). The potential on VPSB (the voltage MAP) is supplied from the load transistors MP2 and MP4 to the gate of the storage transistor MCN1 (i.e. to the node B).
Since the gate potential of the load transistor MP2 (i.e. the node T) is set to 0.5 V, accordingly this transistor is turned ON if the potential of VPSB is greater than or equal to 0.5 V+Vthp, and potential comes to be supplied to the node B. On the other hand, since the gate potential PREB of the load transistor MP4 is set to 0 V, accordingly this transistor is turned ON if the potential of VPSB is greater than or equal to Vthp, so that it becomes possible for potential to be supplied to the node B. Vthp is the threshold voltage of the P-type MOS transistors shown by MP1 through MP4, and is around 0.7 V with a standard CMOS process, and it becomes possible to decide upon the threshold voltage of the transistor MCN1 (the gate voltage which is necessary for some fixed electrical current to flow) in a voltage range greater than or equal to this. When measuring the threshold voltage on the side of the transistor MCN1, the gate potential of the transistor MCN2 (the node T) is set to 0.5 V, in order for the gate potential of the transistor MCN1 (the node B) not to be pulled down by leakage electrical current on the side of the transistor MCN2.
While FIG. 33 shows the voltage application condition for the case of measuring the threshold voltage of the storage transistor MCN1, it is also possible to measure the threshold voltage of the other storage transistor MCN2, simply by controlling the bit lines BLT and BLB, and the pre-charge signals PRET and PREB, in the converse manner.
It should be understood that although, in order to perform this threshold value measurement, the pre-charge line is divided into PRET and PREB, if this measurement is not to be performed, it would be acceptable to provide a single common pre-charge line both to the true side and to the bar side.

Embodiment Three

A non-volatile storage device according to a third embodiment of the present invention, and a semiconductor integrated circuit device which incorporates it, will now be explained with reference to FIGS. 34 through 51.
FIG. 34 is a circuit diagram of a single non-volatile memory cell of this non-volatile storage device. This memory cell is a non-volatile memory cell of an SL divided type eight transistor structure. This memory cell comprises a flip-flop in which an inverter (a true side inverter) in which a P-type MOS transistor MP1 and an N-type MOS transistor MCN1 are connected in series, and an inverter (a bar side inverter) in which a P-type MOS transistor MP2 and an N-type MOS transistor MCN2 are connected in series, are static latch connected. Among these, the P-type MOS transistors MP1 and MP2 are termed load transistors, while the N-type MOS transistors MCN1 and MCN2 are termed storage transistors. As explained in FIG. 36 below, the storage transistors MCN1 and MCN2 function as non-volatile elements which can change their threshold values in a non-volatile manner by accumulating and neutralizing electric charges on their side wall portions.
The storage transistor side end portion of the true side inverter, in other words the source of the storage transistor MCN1, is connected to a true side source line SLT. Furthermore, the storage transistor side end portion of the bar side inverter, in other words the source of the storage transistor MCN2, is connected to a bar side source line SLB. And the load transistor side end portions of both of these inverters, in other words the sources of the load transistors MP1 and MP2, are connected to a line VPS.
In this flip-flop, the inverter in which the load transistor MP1 and the storage transistor MCN1 are connected in series functions as a storage unit on the true side, while the inverter in which the load transistor MP2 and the storage transistor MCN2 are connected in series functions as a storage unit on the bar side. The connection portion between the load transistor MP1 and the storage transistor MCN1 is a node T, while the connection portion between the load transistor MP2 and the storage transistor MCN2 is a node B. When the node T is at high potential and the node B is at low potential, the stored contents is “0”; while, when the node T is at low potential and the node B is at high potential, the stored contents is “1”.
The node T is connected to a bit line BLT (Bit Line-True) via a transfer gate MN1, while the node B is connected to a bit line BLB (Bit Line-Bar) via a transfer gate MN2. The transfer gates MN1 and MN2 are N-type MOS transistors, and a common word line WL is connected to both of these gates.
Furthermore a P-type MOS transistor MP3, which is a transistor for pre-charge, is connected in parallel with the load transistor MP1, in other words between the node T and VPS. Moreover a P-type MOS transistor MP4, which also is a transistor for pre-charge, is connected in parallel with the load transistor MP2, in other words between the node B and VPS. A T-side pre-charge line PRET is connected to the gate of the P-type MOS transistor MP3, while a B-side pre-charge line PREB is connected to the gate of the P-type MOS transistor MP4. Furthermore, all of the P-type MOS transistors MP1 through MP4 are formed within the same N well, and the potential of this N well is controlled by a signal VPM.
FIG. 35 is a figure showing a structure in which a number of the non-volatile cells shown in FIG. 34 are arranged in an array. In this memory cell array, the non-volatile memory cells of FIG. 34 are arranged in an X,Y matrix array. And one word line WL is provided for each row (the rows are arranged along the Y axis direction), while one bit line BLT and one bit line BLB are provided for each column (the columns are arranged along the X axis direction). Each of these word lines WL and bit lines BLT and BLB is controlled independently. On the other hand, the other signal lines (PREB, PRET, VPS, VPM, SLT, and SLB) are provided in common for all of the memory cells, and are controlled all together for the entire memory cell array.
FIG. 36 shows the structure of one of these storage transistors MCN1 (or MCN2), and its operation during writing. Furthermore, FIG. 37 is a figure showing the state of voltage application when writing to this storage transistor MCN1.
In FIG. 36, a P-type well 104 of depth 0.8 μm and average boron density 2·10¹⁷cm⁻³is formed upon a surface region of a P-type silicon substrate 101 of resistivity 10 Ω·cm. The two separate storage transistors MCN1 and MCN2 are defined by a plurality of trenches (element separators) 102 of depth 250 nm on this P-type well 104. In this figure, only one of the storage transistors (MCN1) is shown.
This storage transistor is an N-channel type transistor, and comprises, on the surface region of the P-type well 104, a drain 109 and a source 115 which are formed adjacent to the trenches 102 on the two sides, and a drain extension 107 which is formed in a region adjacent to the drain 109. The drain 109 and the source 115 are formed with an average arsenic density of 1·10²⁰cm⁻³, and the drain extension 107 is formed with an average arsenic density of 5·10¹⁸cm⁻³.
Furthermore, upon the substrate in the channel region, which is the region of the surface of the P-type well 104 between the drain 109 and the source 115, there are formed a gate oxide layer 105 of thickness 5 nm, and a gate electrode 106 which is made from a polysilicon layer of thickness 200 nm and having a phosphorus density of 2·10²⁰cm⁻³. Furthermore, on both sides of this gate oxide layer 105 and gate electrode 106, there are formed side spacers 108 which are made as insulating layers of thickness 50 nm. It should be understood that the side spacer 108S on the source side is exposed to the channel region of the substrate, since no extension region is formed around the periphery of the source 115.
Furthermore, within the region of the P-type well 104, a P-type diffusion layer 111 having an average boron density of 1·10²⁰cm⁻³, which is an electrode for grounding this P-type well, is formed in a region which is separated from the storage transistor described above by one of the trenches 102.
With this storage transistor, the threshold voltage can be elevated by injecting carriers into the side spacer 108S on the source side. Furthermore, as will be explained with reference to FIG. 38, it is also possible to return the threshold voltage to its initial state by extracting the carriers which have been injected into the side spacer 108S. In this way, this storage transistor stores data in a non-volatile manner.
It should be understood that, although the initial threshold voltage of this storage transistor is 1.2 V, the variation is great since it is a transistor of a distinctive structure, and accordingly, from the point of view of reliability, it is not possible to utilize such a storage transistor singly as a storage element. Because of this, in this embodiment, the memory cell is built with the flip-flop structure shown in FIG. 34.
In FIG. 36, the write operation is an operation of injecting channel hot electrons HE into the aforementioned side spacer 108S by applying 0 V to the drain line VD and by applying a positive voltage (for example 6 V) which is less than the junction withstand voltage to the source line VS. Due to this injection of channel hot electrons, the threshold voltage is elevated by trapped electrons, and the system transmits to the write state.
In order to bring the storage transistor MCN1 to this potential configuration, voltages are applied to the memory cell in the condition shown in FIG. 37. This figure shows the state when the threshold voltage of the storage transistor MCN1 is elevated and the data “0” is being written. In the condition with PREB, PRET, VPS, VPM, SLT, and SLB set to 6 V, the word line WL is set to Vcc, and the bit line BLT is set to 0 V while BLB is set to Vcc. By putting ON the transfer gate MN1, which is a N-type MOS transistor, along with the node T (the drain of the storage transistor MCN1) being brought to almost the same potential as the bit line BLT (0 V), since, due to this, the load transistor MP2 is turned ON, accordingly the node B (the gate of the storage transistor MCN1) is brought to almost the same potential as VPSB (6 V). Due to this, the voltage application condition of the storage transistor MCN1 becomes the same as in FIG. 36. At this time, electrical currents of around 300 μA flow in the transfer gate MN1 and in the storage transistor MCN1, and the threshold voltage of the storage transistor MCN1 is elevated to Vth2.
Furthermore, when writing the data “1”, while the threshold voltage on the side of the storage transistor MCN2 becomes elevated, the other conditions are the same as when writing the data “0”, with only the voltage settings for BLT=Vcc and BLB=0 V being reversed.
It should be understood that although, in the embodiment described above, 6V was applied to the gate of the transistor MCN1 (the node B), and 6V was also applied to the source of the transistor MCN1 (the source line SL), it would also be acceptable for these voltages to be different voltages.
FIGS. 38 and 39 are figures showing the conditions in which erase voltages are applied to the storage transistors MCN1 (MCN2). Referring to FIG. 38, the erase operation is an operation in which a positive voltage less than the junction withstand voltage (for example 9 V) is applied to the source line VS, and 0 V is applied to the gate line VG and to the drain line VD, and thereby an avalanche of hot holes HH from the source electrode 115 is created, and these are injected into the side spacer 108S. Due to this, the electrons in the side spacer 108S trapped by the above described write operation are neutralized, and the data which was written is erased by decreasing the threshold voltage.
In the actual memory cell, these voltages are applied under the conditions shown in FIG. 39. Fundamentally, the erase operation is performed all at once for all the memory cells in the memory array shown in FIG. 35. Under the conditions that PREB, PRET, and VPM are set to Vcc, VPS is set to 0 V, and SLT and SLB are set to 9 V, the word lines WL are brought to Vcc and the bit lines BLT and BLB are brought to 0 V. By turning ON the transfer gates MN1 and MN2, which are N-type MOS transistors, the node T and the node B are brought to 0 V, and the voltage application conditions of the storage transistors MCN1 and MCN2 are brought to be the same as those shown in FIG. 38.
Since, in this manner, with the voltage application conditions during writing as shown in FIGS. 36 and 37, and with the voltage application conditions during erasure as shown in FIGS. 38 and 39, the design is arranged so that these operations are performed with the word lines and the bit lines, for which independent control is required for each memory cell, being brought to 0 V or to Vcc, in other words so that the application of high voltage is unnecessary, accordingly it becomes unnecessary to use transistors of high withstand voltage in the control circuitry for the word lines and the bit lines, and thereby it is possible to anticipate the use of transistors of higher performance than in the prior art, whereby the speed of the reading out operation may be increased.
FIGS. 40A to 40C are figures for explanation of the method of threshold value control for the storage transistors MCN1 and MCN2 by the above described write operation, in other words of the method of setting data into this non-volatile memory cell. Here, the data is “1” when the threshold voltage of the storage transistor MCN1 is in the low state (ON) and moreover the threshold voltage of the storage transistor MCN2 is in the high state (OFF); while the data is “0” when the threshold voltage of the storage transistor MCN1 is in the high state (OFF) and moreover the threshold voltage of the storage transistor MCN2 is in the low state (ON).
FIG. 40A shows the case before data setting, in other words when the threshold voltages of both of the storage transistors MCN1 and MCN2 are in their initial state Vth0. Even in this case, due to the procedure shown in FIG. 41 or FIG. 43, the state of this non-volatile memory cell is determined as being data “1”.
FIG. 40B shows the threshold voltages when this non-volatile memory cell is set to the data “0”. The writing of the data “0” is implemented by raising the threshold voltage of the storage transistor MCN1 from its initial state of FIG. 40A to Vth2 (where Vth2>Vth0).
And FIG. 40C shows the threshold voltages when this non-volatile memory cell is set to the data “1”. The writing of the data “1” is implemented by raising the threshold voltage of the storage transistor MCN2 from its initial state of FIG. 40A to Vth2 (where Vth2>Vth0).
When the erase operation explained with reference to FIGS. 38 and 39 is performed, the system returns to the state shown in FIG. 40A, even though the threshold voltages are controlled as shown in FIGS. 40B and 40C.
Since, in this manner, the structure is arranged so that, even though the threshold voltages of the storage transistors MCN1 and MCN2 have once been raised, it is still possible again to lower them to the initial state Vth0, and furthermore since it is possible forcibly to determine the storage transistors MCN1 and MCN2 to the data “1”, even if they are both in the initial state Vth0, accordingly, even though requests for rewriting of the data have been issued a number of times, it is still possible to obtain a sufficient margin for reading out, which is the difference between the threshold voltages on the true side (the storage transistor MCN1) and on the bar side (the storage transistor MCN2).
By the writing operation shown in FIGS. 36 and 37, it is possible to perform writing to either one of the storage transistors MCN1 and MCN2, and it is possible to write the data “1” or “0” into the memory cell. On the other hand, if the memory cell is in its initial state, in other words if writing has not been performed to either one of the storage transistors MCN1 and MCN2 so that the threshold voltages of both of the storage transistors MCN1 and MCN2 are in their initial state Vth0, then generally the contents stored in the memory cell is indeterminate. However, by applying voltages to the memory cell in its initial state according to the following procedure, it is possible to determine the storage contents of this memory cell to either “1” or “0”.
FIG. 41 is a figure for explanation of the voltage application procedure for determining the data of the non-volatile memory cell of the present invention. This procedure is an operation which, if the storage contents of this non-volatile memory cell is “1” or “0”, can set the state of the memory cell (the flip-flop) according to this storage contents, and which can fix the data forcibly at “1” if the memory cell is in its initial state. If, in a memory array in which a plurality of memory cells are arranged, memory cells whose storage contents is “1” and “0” and memory cells in the initial state are mixed together, then, by performing this procedure all at once for all of the memory cells in the memory array, for those memory cells whose storage contents is “1” and “0”, the states of their flip-flops are set according to their storage contents, while, for those memory cells which are in the initial state, the states of their flip-flops are forcibly set to “1”. This procedure is executed during starting up of the memory.
The procedure shown in FIG. 41 proceeds as follows. In the condition with the word line WL and the bit lines BLT and BLB set to 0 V, and with VPS and VPM set to Vcc, first, the source potentials SLT and SLB are raised from 0 V to Vcc at the time instant T0, so that the storage transistors MCN1 and MCN2 are both turned OFF. Then, at the time instant T1, the pre-charge transistors MP3 and MP4 are turned ON by dropping the pre-charge control signals PRET and PREB from Vcc to 0 V, and the node T and the node B are charged to Vcc.
At the time instant T2, PRET and PREB return to Vcc, and, from the time instant T3, the true side source potential SLT slowly drops towards 0 V. Moreover, at a time instant T4 which is more delayed than the time instant T3, the bar side source potential SLB slowly drops towards 0 V. At this time, control is performed so that SLB−SLT=ΔVs (for example ΔVs=0.2 V). Due to this, the voltage between the source and the drain of the storage transistor MCN2 comes to be controlled to be just ΔVs higher than the voltage between the source and the drain of the storage transistor MCN1, and thereby it is possible to make the apparent threshold voltage on the side of the storage transistor MCN2 higher by just the amount ΔVs.
At this time, if the storage contents of the memory cell is “1”, then, since the transistor MCN1 whose threshold voltage is the lower goes to ON before the transistor MCN2, accordingly the node T is pulled down to 0 V first while the node B goes to Vcc, so that the data of the flip-flop is determined at “1”. On the other hand, if the storage contents of the memory cell is “0”, then, since likewise the threshold voltage of the transistor MCN2 is lower than the threshold voltage of the transistor MCN1 (even though its threshold voltage is elevated by just the amount ΔVs) so that the storage transistor MCN2 goes to ON first, accordingly the node B is pulled down to 0 V first while the node T goes to Vcc, so that the data of the flip-flop is determined at “0”.
Furthermore, if the storage contents of the memory cell is “indeterminate”, in other words if the threshold voltages of both the transistors MCN1 and MCN2 are Vth0, then, since the apparent threshold voltage of the storage transistor MCN1 is just ΔVs lower than the threshold voltage of the storage transistor MCN2, accordingly the transistor MCN1 goes ON before the transistor MCN2, and the node T is pulled down to 0 V first while the node B goes to Vcc, so that the data of the flip-flop is determined at “1”.
Since often a memory cell in which the threshold voltages of MCN1 and MCN2 are both Vth0 is a cell to which, up till now, neither writing nor rewriting has been performed, and since, with this type of memory cell, the transistors do not deteriorate along with rewriting, accordingly, with regard to the setting of ΔV, it is sufficient only to pay consideration to variation of the initial threshold voltages of the transistors. Thus it is considered that, for example, around 0.2 V is sufficient.
The voltage application procedure shown in FIG. 41 is the procedure when determining the memory cell in its initial state to the data “1”, but it is also possible to determine the memory cell in its initial state to the data “0” by, in the source potential control of the storage transistors, reversing the potential relationship of SLT and SLB.
FIGS. 42A to 42C are figures for explanation of the margin for data determination when performing the voltage application procedure shown in FIG. 41. In the initial state in which the threshold voltages of the transistors MCN1 and MCN2 are both Vth0, as previously described, by making the source potential SLB on the bar side be just ΔVs higher than the source potential SLT on the true side, the apparent threshold voltage on the MCN2 side is increased by just ΔVs, so that the data is forcibly recognized as “1” (FIG. 42A). Although, for a memory cell in which the data “0” has already been written, the margin comes to be decreased by just ΔVs, suppose that Vth2−Vth0=1 V and ΔVs=0.2 V: then the margin becomes 0.8 V (FIG. 42B). And conversely, for a memory cell in which the data “1” has already been written, by increasing the margin by just ΔVs, suppose that Vth2−Vth0=1 V and ΔVs=0.2 V: then the margin becomes 1.2 V (FIG. 42C).
FIG. 43 is a figure for explanation of the procedure for voltage application for determining the data for a memory cell which is in its initial state. In other words, while the procedure shown in FIG. 43 is a procedure which, for a memory array in which memory cells whose storage contents are “1” and “0” and memory cells which are in their initial state are mixed together, determines the data of these memory cells all at once, by contrast, the procedure shown in FIG. 43 is a procedure which is only effective for a memory cell which is in its initial state, and which reliably determines the data of such a memory cell in its initial state as being either “1” or “0”.
By executing this procedure, if memory cells to which data has already been written and memory cells which are in their initial state are mixed together, individually for those memory cells which are in their initial states, or by executing this procedure all at once for a memory array of which all of the memory cells are in their initial state, it is possible to determine the data of those memory cells as “1” or “0”. In the following explanation, the procedure will be shown for forcibly determining the storage contents of a memory cell in its initial state to “1”.
Under the condition that the word line WL and the bit lines BLT and BLB are set to 0 V, and that VPS and VPM are set to Vcc, first at the time instant T0 the source potential SLB on the bar side is raised from 0 V to Vcc and the storage transistor MCN2 is turned OFF. Then, at the time instant T1, by dropping the pre-charge control signal PREB from Vcc to 0 V, the node B is charged up to Vcc by the pre-charge transistor MP4. At the time instant T2 PREB returns to Vcc, and then, at the time instant T3, SLB is returned from Vcc to 0 V. Due to this, by the transistor MCN1 whose gate potential is the higher going to ON before the transistor MCN2, in the state in which the node B is held at Vcc, the node T is pulled down to 0 V, so that the data in this flip-flop is determined as being “1”.
The procedure shown in FIG. 43 is a procedure for forcibly determining the storage contents of the memory cell to “1”, but it would also be possible forcibly to determine the storage contents to “0”. In other words, by raising the true side source potential SLT from 0 V to Vcc at the time instant T1, instead of the bar side source potential SLB, and by dropping the gate potential PRET of the pre-charge transistor MP3 from Vcc to 0 V at the time instant T1, instead of the gate potential PREB of the pre-charge transistor MP4, the node B is set to 0 V, so that the data is determined at “0”.
By executing the procedure of FIG. 41 or the procedure of FIG. 43, even if memory cells are present in their initial state in which both of the two storage transistors MCN1 and MCN2 are in their initial state of Vth0, it is possible to determine their storage contents uniquely at “1” or at “0”.
If this non-volatile memory cell is used instead of a fuse, and if its storage contents are uniquely determined to “1”, then, still in the initial state, it is possible to implement the state before cutting the fuse (corresponding to the data “1”).
FIGS. 44 and 45 are figures showing the voltage application conditions when reading out from this non-volatile memory cell. Referring to FIG. 44, the operation of reading out is performed by reading out the voltage on the drain line VD, when the source line VS is set to 0 V and Vcc has been applied to the gate line VG. In the actual memory cell, the voltages are applied in the conditions shown in FIG. 45. This read out operation is the same as the operation of reading out an SRAM, and employs a differential type sense amp. PREB, PRET, VPS, and VPM are set to Vcc and SLT and SLB are set to 0 V, and the change in the bit lines BLT and BLB, which corresponds to the data in this flip-flop, is read out by the differential type sense amp, under the condition that the word line WL is set to Vcc. When BLT is at a low voltage (0 V) and moreover BLB is at a high voltage (Vcc), then the data is “1”; while, when BLT is at a high voltage (Vcc) and moreover BLB is at a low voltage (0 V), then the data is “0”.
While the data determination procedure shown in FIG. 43 shows a method which, even if memory cells in their initial state (i.e. memory cells for which both of the storage transistors MCN1 and MCN2 are in their initial state (with threshold voltage=Vth0)) are present, can forcibly determine them to “1”, if no such memory cells in their initial state are present, or if it will be no problem even if the data values of the memory cells in their initial state are indeterminate, then it is possible to employ the determination procedure shown in FIGS. 46 through 48. With this procedure, no decrease (refer to FIG. 43) of the determination margin by ΔVs takes place, as in the procedure shown in FIG. 41.
The procedure shown in FIG. 46 is conducted as follows. Under the condition that the word line WL and the bit lines BLT and BLB are set to 0 V and VPS and VPM are set to Vcc, first at the time instant T0 the source potentials SLT and SLB are raised from 0 V to Vcc, and the storage transistors MCN1 and MCN2 are turned OFF. In this state, by lowering the pre-charge control signals PRET and PREB from Vcc to 0 V at the time instant T1, the pre-charge transistors MP3 and MP4 are turned ON, and the node T and the node B are pre-charged to Vcc. After PRET and PREB return to Vcc at the time instant T2 and the pre-charging has been completed, from the time instant T3, the source potentials SLT and SLB slowly sink towards 0 V. At this time, if the data is “1”, then, since the transistor MCN1 whose threshold voltage is the lower goes to ON first before the transistor MCN2, accordingly the node T is pulled down to 0 V first while the node B goes to Vcc, so that the data of the flip-flop is determined at “1”. On the other hand, if the data is “0”, then, since the transistor MCN2 whose threshold voltage is the lower goes to ON first before the transistor MCN1, accordingly the node B is pulled down to 0 V first while the node T goes to Vcc, so that the data of the flip-flop is determined at “0”.
And the procedure shown in FIG. 47 is conducted as follows. The word line WL and the bit lines BLT and BLB are set to 0 V, and VPS and VPM are set to Vcc, and moreover the source potentials SLT and SLB are fixed at 0 V. Under these conditions, by dropping the pre-charge control signals PRET and PREB from Vcc to 0 V at the time instant T0, the pre-charge transistors MP3 and MP4 are turned ON, and the node T and the node B are pre-charged. With regard to the pre-charge level of the node T and the node B, the current flow amount which flows as DC in each of the transistors MP1, MP3, and MCN1, or in each of the transistors MP2, MP4, and MCN2, now becomes constant and stabilizes. At this time, if the data is “1”, then, since the current drive force of the transistor MCN1 whose threshold voltage is the lower is greater than that of MCN2, accordingly the node T goes to a lower potential than the node B, and, after completion of the pre-charge at the time instant T1, the data of the flip-flop is determined at “1”. On the other hand, if the data is “0”, then, since the current drive force of the transistor MCN2 whose threshold voltage is the lower is greater than that of MCN1, accordingly the node B goes to a lower potential than the node T, and, after completion of the pre-charge at the time instant T1, the data of the flip-flop is determined at “0”.
To compare this determination procedure with the determination procedure shown in FIG. 46, although it has a demerit from the point of view of consumption of electrical power since a feedthrough DC current flows in the flip-flop during the pre-charge process, nevertheless it has the merit from the point of view of simplicity and ease of control that it is possible to control the source line potentials SLT and SLB by fixing them at 0 V.
Moreover, the procedure shown in FIG. 48 is conducted as follows. In this data determination procedure, an important point is that the pre-charge voltages of the node T and the node B are supplied from the bit line side. Along with fixing PRET and PREB at Vcc, the word line WL and the bit lines BLT and BLB are set to Vcc, and, by setting the potential VPS to 0 V, thus cutting off the transistors MP1 through MP4, the node T and the node B are each pre-charged to Vcc−Vthn via the transfer gates MN1 and MN2 from the bit lines BLT and BLB which have been charged up to Vcc. Here Vthn is the threshold voltage of the transistors MN1 and MN2. At the time instant T0 the word line WL is dropped from Vcc to 0 V, and, with the node T and the node B made floating, the data of the flip-flop is determined by the difference between the amounts of electrical charge discharged from the transistors MCN1 and MCN2. If the data is “1”, then, since the current drive force of the transistor MCN1 whose threshold voltage is the lower is greater than that of MCN2, accordingly the node T goes to a lower potential than the node B, and, after the potential VPS has risen from 0 V to Vcc at the time instant T1, the data of this non-volatile memory cell is determined at “1”. On the other hand, if the data is “0”, then, since the current drive amount of the transistor MCN2 whose threshold voltage is the lower is greater than that of MCN1, accordingly the node B goes to a lower potential than the node T, and, after the potential VPS has risen from 0 V to Vcc at the time instant T1, the data of this non-volatile memory cell is determined at “0”.

Variants of Embodiment Three

With this eight transistor SL divided type non-volatile memory cell of the third embodiment, in the state in which writing has not been performed, in other words in the state in is which the threshold voltages of the storage transistors MCN1 and MCN2 are both low, the data cannot be determined by performing the raising processing (i.e. the data determination processing) shown in FIGS. 46 through 48. Thus, as shown in FIGS. 49 and 50, by unbalancing the circuit structure on the true side (the storage transistor MCN1) and on the bar side (the storage transistor MCN2), it is possible to determine the data as being one of “0” or “1” upon raising.
The non-volatile memory cell shown in FIG. 49 is one in which the channel widths of the two load transistors MP1 and MP2 are unbalanced. In this example, the channel width of the load transistor MP1 (shown by the thick lines) is made to be twice the channel width of the load transistor MP2, so that the resistance value of the load transistor MP1 when it is ON becomes about ½ of the resistance value of the load transistor MP2 when it is ON.
When the power supply to the memory cell having this structure is turned ON in the state in which neither one of the storage transistors MCN1 and MCN2 has been written, the potential at the node T rises faster than that at the node B, and the load transistor MP1 and the storage transistor MCN2 go into the ON state, while the load transistor MP2 and the storage transistor MCN1 go into the OFF state; in other words, the data stabilizes at “0”.
It should be understood that, instead of unbalancing the channel widths as described above, it would also be acceptable to unbalance the channel lengths. Furthermore, it would also be acceptable for the load transistor whose channel width or channel length is changed to be either MP1 or MP2. Moreover, it would also be acceptable to perform this unbalancing by changing the channel width or the channel length of one of the storage transistors MCN1 or MCN2.
On the other hand, the non-volatile memory cell shown in FIG. 50 is one in which a capacitor is connected to each of the two nodes T and B of the flip-flop. A capacitor C1 is connected between the node T and the power supply line Vcc, while a capacitor C2 is connected between the node B and ground. The capacitances of these capacitors may be, for example, around 50 fF.
Due to this, when the power supply is turned ON in the state in which neither writing nor erasure of either of the two storage transistors MCN1 and MCN2 has been performed, since the potential of the node T rises quickly directly after the power supply has been turned ON while the potential of the node B rises more slowly, accordingly the system stabilizes with the load transistor MP1 and the storage transistor MCN2 in the ON state and the load transistor MP2 and the storage transistor MCN1 in the OFF state, in other words at the data “0”.
Although, in the example shown in FIG. 50, capacitors were connected to both the node T and to the node B, it would also be possible to provide a non-volatile memory cell of an asymmetric circuit structure, with a capacitor being connected to only one of the nodes.
—The Threshold Voltage Measurement Method in the Third Embodiment—
It should be understood that, with this non-volatile memory cell of an eight transistor structure, it is possible to measure the threshold voltages of the storage transistors by establishing a potential arrangement as, for example, shown in FIG. 51. By measurement of the threshold voltages by this method, it becomes possible to perform evaluation of the variation of the threshold voltages in the initial state, and evaluation of the amount of change of the threshold voltages during a write operation or an erase operation and of the high temperature storage characteristics of the threshold voltages after rewriting and so on.
In FIG. 51, the case is shown of deciding upon the threshold voltage of the storage transistor MCN1. In the state in which the source potential SLT of the storage transistor MCN1 which is to be tested is set to =0 V, 1 V is supplied from the bit line BLT via the transfer gate MN1 to the drain of the storage transistor MCN1 (the node T). And the potential on VPS (the voltage MAP) is supplied from the load transistors MP2 and MP4 to the gate of the storage transistor MCN1 (i.e. to the node B).
Since the gate potential of the load transistor MP2 (i.e. the node T) is set to 1 V, accordingly this transistor is turned ON if the potential of VPSB is greater than or equal to 1 V+Vthp, and potential comes to be supplied to the node B. On the other hand, since the gate potential PREB of the load transistor MP4 is set to 0 V, accordingly this transistor is turned ON if the potential of VPSB is greater than or equal to Vthp, so that it becomes possible for potential to be supplied to the node B. Vthp is the threshold voltage of the P-type MOS transistors shown by MP1 through MP4, and is around 0.7 V with a standard CMOS process. Accordingly, it becomes possible to decide upon the threshold voltage of the transistor MCN1 (the gate voltage which is necessary for some fixed electrical current to flow) in a voltage range greater than or equal to this. When measuring the threshold voltage on the side of the transistor MCN1, the source potential of the transistor MCN2 and the MAP voltage and the bit lines are set to floating, in order to eliminate unnecessary leak-pass on the side of the transistor MCN2.
While FIG. 51 shows the voltage application condition for the case of measuring the threshold voltage of the storage transistor MCN1, it is also possible to measure the threshold voltage of the other storage transistor MCN2, simply by controlling the bit lines BLT and BLB, by controlling the source potentials SLT and SLB, and by controlling the pre-charge signals PRET and PREB, in the converse manner.
It should be understood that although, in order to perform this threshold value measurement, the pre-charge line is divided into PRET and PREB, if this measurement is not to be performed, it would be acceptable to provide a single common pre-charge line both to the true side and to the bar side.
—The Layout of the Second and Third Embodiments Upon the Semiconductor Substrate—
FIG. 52 shows the layout upon the semiconductor substrate of these memory cells according to the second and the third embodiments.
(A) of this figure shows the layout of the active regions and the gate electrodes upon the surface of the substrate. And (B) of this figure shows the second and third layers of metal wiring. As shown in (A) of this figure, the storage transistors on the true side and on the bar side are laid out in the same direction, so that it is arranged for the characteristics still to be mutually the same, even if an error occurs due to an impurity intruding diagonally.
Furthermore, as shown in (C) of this figure, the VPS lines and the source lines (SL) are divided, in this layout, into lines on the true side and lines on the bar side, so that, by shorting together either pair of these (making them common), it is possible to build a structure of the VPS divided type as in the second embodiment, or a structure of the SL divided type as in the third embodiment. In this manner, by employing the layout of this figure, while keeping the processing as far as the second layer of metal wiring the same, it is possible to manufacture either a non-volatile semiconductor memory cell of the VPS divided type of the second embodiment, or a non-volatile semiconductor memory cell of the SL divided type of the third embodiment, simply by changing only the third layer of metal wiring.

Embodiments Four and Five

FIG. 53 is a circuit block diagram of an RFID chip to which a non-volatile memory cell as described above is provided. An antenna L configured externally to this chip is connected to pads P1 and P2, in order to receive an RF signal which is generated by an external reader. A power supply capacitor CT of capacitance 120 pF, a voltage clamp circuit VOLTAGE CLAMP, a power supply modulator MODULATOR, and a bridge rectifier BRIDGE RECTIFIER are connected between these pads P1 and P2. A power supply stabilization capacitor CF is connected to the output of the bridge rectifier, and, by feeding back to the voltage clamp circuit a control signal from a voltage regulator REGULATOR for detecting the output voltage, stabilization of the power supply voltage may be anticipated. Furthermore, a Vcc detection circuit VCC DETECTOR which generates an internal power supply voltage (Vcc) and a voltage generation circuit VPP GENERATOR which generates various voltages other than Vcc are connected to the output of the bridge rectifier. Moreover, a circuit MODE SELECTOR which detects the operational mode included in the RF signal received by the bridge rectifier, a clock signal detection circuit CLOCK EXTRACTOR, and a circuit DATA MODULATOR which extracts write data to a module EEPROM of a non-volatile storage device, are provided to the output of the bridge rectifier. And a controller CONTROLLER controls the operation of the module EEPROM of the non-volatile storage device which receives the operation mode data.
In the module EEPROM of the non-volatile storage device which is mounted to this RFID chip, there are written an ID number for authenticating the chip, an address for delivery service, goods information which is an alternative to a bar code (price, day of production, place of production, producer, component information, and so on), information required for an air cargo tag (name of carrier, name of owner, place of loading, destination, and so on), or the like.
FIG. 54 is a schematic plan view of a system LSI chip which is an example of a semiconductor integrated circuit device to which a non-volatile memory cell according to the present invention is mounted. Although the system LSI shown in this figure is not particularly limited, it includes a semiconductor substrate around the border of which are arranged a large number of external connection electrodes 120 such as bonding pads or the like, with an external input and output circuit 121 and an analog input and output circuit 122 being provided on the inside thereof. This external input and output circuit 121 and analog input and output circuit 122 take an external power supply, whose level such as 3.3 V is relatively high, as a supply of operating power. And a level shift circuit 123 drops this external power supply to an internal power supply voltage such as 1.8 V.
In the interior of the level shift circuit 123, there are provided a static random access memory (SRAM) 124, a central processing device (CPU) 125, a cache memory (CACH) 126, a logic circuit (LOGIC) 127, a phase locked loop circuit (PLL) 128, an analog-digital conversion circuit (ADC) 129, a digital-analog conversion circuit (DAC) 130, and a system controller (SYSC) 131. Each of the units designated as 132, 133, and 134 is an electrically erasable and writable non-volatile memory (EEPROM), and is a non-volatile storage device according to the present invention having an appropriate predetermined capacity.
The non-volatile memory 132 is used for storing repair information for the SRAM 124 (i.e. control information for replacing defective memory cells with redundant memory cells). In other words, this semiconductor integrated circuit device includes circuits which may require repair (memory cells within the SRAM 124, if they become defective) and redundancy circuits (redundant memory cells within the SRAM 124) which can serve as alternatives to those circuits which may require repair; and, moreover, includes the non-volatile memory 132 which stores repair information. This repair information is information which specifies which circuits requiring repair may be replaced by which redundancy circuits.
Moreover, this semiconductor integrated circuit device includes an analog circuit and a constant trimming circuit which adjusts one or more circuit constants of this analog circuit. In this structure, the non-volatile memory 133 is used for storing information in this constant trimming circuit for specifying these circuit constants. Furthermore, this semiconductor integrated circuit device includes an oscillation circuit and a frequency trimming circuit which adjusts the oscillation frequency of the oscillation circuit. In this structure, the non-volatile memory 133 is used for storing information within this frequency trimming circuit for specifying this oscillation frequency. Yet further, this semiconductor integrated circuit device includes a reference voltage generation circuit and a voltage trimming circuit which adjusts the reference voltage generated by the reference voltage generation circuit. In this structure, the non-volatile memory 133 is used for storing information within this voltage trimming circuit for specifying this reference voltage. Even further, this semiconductor integrated circuit device includes a security circuit for identifying the chip. In this structure, the non-volatile memory 133 is used for storing information within this security circuit for identifying the chip.
Finally, the non-volatile memory 134 is endowed with 256 bits of memory capacity, and is used for storing ID information for this chip, operational mode information for this chip, and other appropriate data.

Claims

1. A non-volatile semiconductor storage device, comprising:

a flip-flop in which two inverters, each consisting of a load transistor and a storage transistor connected in series, are cross-connected;

two gate transistors, each respectively connected to a node of said flip-flop on a side thereof;

two bit lines, each of which is connected to a respective one of said two gate transistors, and which are controlled to a voltage between operating power supply voltage and ground voltage;

a word line which is connected to both of the gate electrodes of said two gate transistors, and which is controlled to a voltage between operating power supply voltage and ground voltage;

a first voltage supply line which is connected to the sources of said storage transistors of said inverters, and to which a predetermined first voltage is supplied during writing and during erasure; and

a second voltage supply line which is connected to the sources of said load transistors of said inverters, and to which a predetermined second voltage is supplied during writing; wherein said storage transistors of said inverters are constituted by storage transistors which can be threshold voltage controlled by injection of electrons into the neighborhood of their gates.

2. A non-volatile semiconductor storage device according to claim 1, wherein each of said storage transistor comprises:

an insulation layer side spacer which is formed at a side portion of its gate electrode; and

a low impurity density region which is formed at a border portion of its drain; and wherein:

during writing, said first voltage supply line applies said first voltage to the sources of said storage transistors, and said second voltage supply line applies said second voltage to the gates of said storage transistors via the sources of said load transistors; and thereby information is written into said storage transistors by channel hot electrons being injected into said insulation layer side spacers; and

during erasure, said first voltage supply line applies said first voltage to the sources of said storage transistors; and thereby information which is stored in said storage transistors is erased by an avalanche of hot holes being injected into said insulation layer side spacers.

3. A non-volatile semiconductor storage device according to claim 1, wherein each said inverter comprises a transistor for pre-charge which is connected in parallel with said load transistor thereof; and

further comprising pre-charge control voltage supply lines which supply pre-charge control voltages to the gates of said transistors for pre-charge; and

wherein said transistors for pre-charge are ON/OFF controlled independently from said load transistors by said pre-charge control voltages supplied by said pre-charge control voltage supply lines.

4. A non-volatile semiconductor storage device according to claim 3, wherein said second voltage supply line is provided separately for each of said inverters.

5. A non-volatile semiconductor storage device according to claim 3, wherein said first voltage supply line is provided separately for each of said inverters.

6. A non-volatile semiconductor storage device according to claim 1, wherein the continuity resistances of the load transistors of said two inverters are unbalanced.

7. A non-volatile semiconductor storage device according to claim 1, wherein the electrostatic capacitances with respect to the power supply voltage lines, or with respect to ground, of said two inverters are unbalanced.

8. A non-volatile semiconductor storage device according to claim 1, wherein the continuity resistances of the storage transistors of said two inverters are unbalanced.

9. A method for determining the state of a non-volatile semiconductor storage device according to claim 3, comprising a step of supplying a pre-charge control voltage selectively to one only of said pre-charge control voltage supply lines, and thereby turning only one but not the other of said transistors for pre-charge of said two inverters ON, thus forcibly determining the state of said flip-flop.

10. A method for determining the state of a non-volatile semiconductor storage device according to claim 4, comprising:

a step of supplying a voltage lower than said first voltage to said first voltage supply line, and thereby raising the source potentials of said storage transistors of said two inverters and turning said storage transistors OFF;

a step of supplying different voltages to said second voltage supply lines, thereby applying pre-charge voltages of different potentials to said nodes via said transistors for pre-charge of said two inverters; and

a step of gradually decreasing said low voltage supplied to said first voltage supply line, thereby gradually lowering the source potentials of said storage transistors of said two inverters.

11. A method for determining the state of a non-volatile semiconductor storage device according to claim 4, comprising:

a step of supplying voltages lower than said first voltage to said first voltage supply lines, and thereby raising the source potentials of said storage transistors of said two inverters and turning said storage transistors OFF;

a step of supplying the same voltage to said second voltage supply lines, thereby applying pre-charge voltages of the same potential to said nodes via said transistors for pre-charge of said two inverters; and

a step of gradually decreasing said low voltages supplied to said first voltage supply lines while maintaining a predetermined potential difference therebetween, thereby gradually lowering the source potentials of said storage transistors of said two inverters while maintaining a predetermined potential difference therebetween.

12. A semiconductor integrated circuit device comprising a non-volatile semiconductor storage device as described in claim 1, a circuit which may require repair comprising a portion which may require repair, and a redundancy circuit which serves as an alternative to said circuit which may require repair; and

wherein said non-volatile semiconductor storage device is a storage circuit which stores repair information specifying which circuit may require repair by said redundancy circuit serving as an alternative thereto.

13. A semiconductor integrated circuit device comprising a non-volatile semiconductor storage device as described in claim 1, an analog circuit, and a constant trimming circuit which adjusts a circuit constant of said analog circuit; and

wherein said non-volatile semiconductor storage device is a storage circuit which stores information in said constant trimming circuit for specifying said circuit constant.

14. A semiconductor integrated circuit device comprising a non-volatile semiconductor storage device as described in claim 1, an oscillation circuit, and a frequency trimming circuit which adjusts the oscillation frequency of said oscillation circuit;

and wherein said non-volatile semiconductor storage device is a storage circuit which stores information in said frequency trimming circuit for specifying said oscillation frequency.

15. A semiconductor integrated circuit device comprising a non-volatile semiconductor storage device as described in claim 1, a reference voltage generation circuit, and a voltage trimming circuit which adjusts the reference voltage generated by said reference voltage generation circuit;

and wherein said non-volatile semiconductor storage device is a storage circuit which stores information in said voltage trimming circuit for specifying said reference voltage.

16. A semiconductor integrated circuit device comprising a non-volatile semiconductor storage device as described in claim 1, and a security circuit which identifies a chip in which said semiconductor integrated circuit device is mounted; and

wherein said non-volatile semiconductor storage device is a storage circuit which stores information in said security circuit for specifying said chip.