JPH09223780A - Nonvolatile semiconductor storage device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a semiconductor storage device of high reliability which allows no uselessness, by arranging a first storage region and a second storage region different in read/write characteristics, and accessing the storage region in response to the reliability of read/write required for the data as the access object. SOLUTION: A memory cell array is divided into a first part connected with word lines WL1-WLk and a second part connected with word lines WL(k+1)-WLm. A first row decoder 32-1 for selecting the word lines WL1-WLk and a second row decoder 32-2 for selecting the word lines WL(k+1)-WLm are individually installed. The first part is a storage region of high reliability wherein a one-bit data storage device is constituted of two cells. The second part is a usual storage region wherein a one-bit storage device is constituted of a cell. Therefore, reliability is improved and increase of the chip area can be restrained to a minimum.

Description

Detailed Description of the Invention

[0001]

The present invention relates to a nonvolatile semiconductor memory device.

[0002]

2. Description of the Related Art A ROM capable of electrically erasing and rewriting stored contents is known as an EEPROM (electrically erasable programmable ROM). This EEPROM is an ultraviolet erasing type EPR.
Compared to OM, data can be erased by an electric signal while mounted on a board. For this reason, EEP
The ROM is widely used for various controls and as a memory card.

FIG. 15 is a sectional view showing the element structure of a typical memory cell in the EEPROM, and FIG. 16 is an equivalent circuit diagram thereof. In FIG. 15, for example, on a P-type substrate 80, N-type diffusion regions 91, 92 and 93 are provided. Between the diffusion regions 91 and 92 on the substrate 80, a floating gate electrode 95 formed of a first polycrystalline silicon layer is provided via an insulating oxide film 94. This floating gate electrode 95 is connected to the thin film portion 9 of the insulating oxide film 94.
4A, it overlaps with the N-type diffusion region 92.
On the floating gate electrode 95, a gate electrode 97 formed of a second-layer polycrystalline silicon layer is provided via an insulating oxide film 96. Between the diffusion regions 92 and 93 on the substrate 80, a gate electrode 99 formed of a first-layer polycrystalline silicon layer is provided via an insulating oxide film 98.

The memory cell shown in FIG. 15 has two transistors 1 and 2. That is, one is the N-type diffusion region 91.
Is a floating gate transistor 2 as a nonvolatile memory element having a source, an N-type diffusion region 92 as a drain, a floating gate electrode 95 as a floating gate, and a gate electrode 97 as a control gate. The other is to use the N-type diffusion region 92 as a source,
This is an enhancement type select transistor 1 having the drain region 93 as a drain and the gate electrode 99 as a gate.
These transistors 1 and 2 are connected in series.
Then, as shown in the equivalent circuit of FIG. 16, the drain and the gate of the transistor 1 are used as the data line DL and the word line WL. The floating gate and the control gate of the floating gate transistor 2 are used as the floating gate FG and the control gate CG, and the source is used as the source S, respectively. The memory cell of FIG. 11 is a 1-bit data storage unit (storage unit) that stores 1-bit data.
Is composed.

Table 1 Operation mode V _WL V _CG V _S V _DL V _FG Floating gate status Erasure H H 0V 0V H Electron injection from drain to floating (“1”) gate Write H 0V H H L Electron emission from floating gate to drain (“0”) drain Write H 0V H 0V L No electron movement (“1”) Read 5V 0V 0V 1V- Table 1 shows the operation modes of the memory cell shown in the equivalent circuit of FIG. In this memory cell,
There are four operation modes: erase, "0" write, "1" write, and read. Hereinafter, these operation modes will be described.

Erase Mode The word line WL and the control gate CG are in the selected state, and the potentials V _WL and V _CG are respectively high potential H (for example, 20 V).
Is applied, and 0 V is applied to the data line DL. At this time, the potential V _FG of the floating gate FG becomes a high potential H (for example, about 12 V) due to capacitive coupling with the control gate CG. In addition, since the selection transistor 1 is turned on and the drain potential of the floating gate transistor 2 is 0V, the tunnel effect of Fowler-Noldheim causes the thin film portion 94A in FIG. Electrons are injected from the drain of the floating gate transistor 2 to the floating gate FG. This operation is called a data erase operation. The data after erasing is set to the “1” level.

In the write modes of the data "0" and "1", the potential V _WL of the word line _WL is set to the high potential H in both modes.
The potential V _CG of the control gate CG becomes 0 V and the potential V _CG of the source S becomes
_S is set to a high potential H (for example, 5 V). Furthermore, "0"
In the write mode, the potential VDL of the data line _DL is set to the high potential H (data input “0”). Floating gate F
G has a low potential L due to capacitive coupling with the control gate CG. In this case, electrons are emitted from the floating gate FG of the floating gate transistor 2 to the drain through the thin film portion 94A by the Fowler-Nordheim tunnel effect. This operation is called a data “0” write operation.

On the other hand, in the "1" write mode, the potential V _DL of the data line DL is set to 0 V (data input "1"). On the other hand, the floating gate FG has almost no potential difference from the control gate CG and becomes 0V. In this case, there is no transfer of electrons. Therefore, if electrons are previously injected into the floating gate FG and the data becomes “1”, the state is maintained. This operation is called a data “1” write operation.

Data Read Mode The potential V _WL of the word line WL is set to 5 V, the potential V _DL of the data line DL is set to about 1 V, and the potential V _CG of the control gate CG is set to 0 V. As a result, ON / OFF of the floating gate transistor 2 is determined according to the type of charge (electrons or holes) stored in the floating gate FG. For example, in a state where electrons are accumulated in the floating gate FG (stored data is at “0” level), the floating gate transistor 2 is turned off. At this time, no cell current flows. On the other hand, when holes are accumulated in the floating gate FG (stored data is at “1” level), the floating gate transistor 2 is turned on, and a cell current flows. Such data reading is performed by a sense amplifier circuit that operates according to the presence or absence of a cell current.

In the memory cell using the floating gate transistor as described above, data written once is ideally semipermanently retained unless data is erased. However, in an actual memory cell, after erasing or writing data, the charge in the floating gate is released with the passage of time, and the stored data is lost. In particular, the charge loss is remarkable in a cell having a defect in an insulating oxide film or the like. In some cases, it may be defective during use.

Generally, as a method of evaluating the retention characteristic of stored data, there is a method of accelerating the time of occurrence of a defect by setting a high temperature state. This is called a high-temperature storage test. FIG.
No. 7 shows that when this high temperature storage test was performed at 300 ° C.
FIG. 4 is a characteristic curve diagram showing a change in a threshold voltage (V _TH ) of a floating gate transistor. The threshold voltage in the initial state is about 1 V as shown by the broken line.

First, electrons are emitted from the floating gate,
A case where data of the “0” level is stored will be described. At this time, the threshold voltage of the floating gate transistor becomes substantially a negative value, for example, -5V. Therefore, current flows even when the potential of the control gate is 0V.

Next, electrons are injected into the floating gate,
A case where data of “1” level is stored will be described. The threshold voltage of the floating gate transistor becomes a substantially high value, for example, + 10V.

At the time of reading data, the control gate potential is set to 0V. Then, whether the data stored in the memory cell is “0” or “1” is determined by determining the operating point of the sense amplifier circuit, that is, the sense potential.
This is performed by setting an appropriate current to flow in the memory cell. This sensing potential is set to about -1 V as shown by a dashed line in the figure.

In FIG. 17, in the cell of "1" data, electrons in the floating gate are emitted with the lapse of time.
Thereby, the threshold voltage decreases over time,
It approaches the initial threshold voltage of 1V. On the other hand,
In the cell of “0” data, electrons are injected into the floating gate with the passage of time. As a result, the threshold voltage increases with time and approaches 1V. In the middle of the time t _N, it passes through -1V a sense potential of the sense amplifier circuit.

FIG. 18 shows a cell current (I c at the time of a high temperature storage test of a memory cell storing “0” level data).
ell) indicates a change. The cell current decreases with time. When the current becomes lower than the sense level current I _S in the current value sense amplifier circuit, the sense amplifier circuit erroneously determines that the data which was originally at “0” level is “1”. The reason that data may be detected by mistake is that
Only memory cells that store "0" level data. The time at which this erroneous data is detected is denoted by t _N
And The time required to reach the time t _N is sufficiently long for a normal memory cell, and there is no problem in practical use.
However, the time until the time t _N is short in a defective memory cell. Therefore, a defect may occur during use of the product. In particular, when erasing and writing are frequently repeated, the insulating oxide film is significantly deteriorated, and defects are likely to occur.

FIG. 19 is a circuit diagram of a typical conventional EEPROM in which a memory cell shown in the equivalent circuit of FIG. 16 is used to form a cell array. Each memory cell M
The control gates of the floating gate transistors 2 of C-11 to MC-mn are connected via the control gate selection transistor 6,
They are connected to control gate selection lines CGL1 to CGLn selected by the column decoders 5-1 to 5-n. Further, the control gate selection transistor 6 in the same memory cell
And the gate of the selection transistor 1 are connected to one of the row lines WL1 to WLm selected by the row decoder 4. The drain of the select transistor 1 in each memory cell is connected to column lines DL1 to DLn. The column lines DL1 to DLn are respectively connected to the column selection transistors 7
Are connected to the bus line 8 via the. The gate of the transistor 7 is connected to the column decoder 5 via the column selection lines CL1 to DLn.
It is connected to the. A data input circuit 9 and a sense amplifier circuit 10 are connected to the bus line 8. The data input circuit 9 outputs “0” or “1” level data according to the write data signal Din input from the outside. The sense amplifier circuit 10 detects the level of the storage data in the selected memory cell MC as “0” or “1”. At the time of the detection, the sense amplifier circuit 10 applies a bias voltage necessary for data reading to the data line DL. That is, the sense amplifier circuit 10 includes a bias circuit.

The data detected by the sense amplifier circuit 10 is input to the data output circuit 12. The read data is output from the data output circuit 12 to the outside. In the EEPROM having such a structure, the probability of random bit-like cell failure due to the above-mentioned defects is 10 ^{3 in} a device having a storage capacity of 64 Kbits.
When erasing and writing are performed about once, it is about 0.1% to 0.2%. For this reason, there is a drawback that practical applications are limited.

FIG. 20 is a circuit diagram of an example of a conventional EEPROM in which the above defect rate is greatly improved. As described above, the defective memory cells randomly occur only in the memory cells storing "0" level data. Therefore, in the EEPROM of FIG. 20, the same data is stored in two memory cells. Even if the “0” data of one memory cell becomes defective, if the other “0” data is normal, normal data is read.

That is, this EEPROM is constructed as follows. The two serial circuits 3A and 3B constitute a 1-bit memory cell (1-bit data storage) MC for storing one data. The series circuits 3A and 3B include the selection transistors 1A and 1B and the floating gate transistor 2
A, 2B. The drains of the select transistors 1A and 1B in the memory cell are connected to column lines DLiA and DLiB (i =
1 to n). The above column line DLi
A and DLiB are connected to bus lines 8A and 8B via column selection transistors 7A and 7B. The bus line 8A,
8B are both connected to the same data input circuit 9, and are connected to the sense amplifier circuits 10A and 10B, respectively. The outputs of both sense amplifier circuits 10A and 10B are input to AND logic circuit 11. The output of the logic circuit 11 is input to the data output circuit 12.

In the EEPROM having such a structure, when one memory cell is selected, two series circuits 3 in the memory cell are selected.
A and 3B are simultaneously selected. Therefore, the possibility of normal operation increases. That is, one of the series circuits is “0”.
Suppose that it became defective. Thereby, the sense amplifier circuit 1
One of the outputs 0A and 10B becomes "1" level. However, it is assumed that the other output is at the normal "0" level. At this time, the output of the logic circuit 11 becomes "0" level. As a result, a normal operation is performed.

The probability of simultaneous occurrence in the two series circuits 3A and 3B as normal memory cells having random defects as described above is very small. Therefore, in the method of providing such two series circuits, the defect occurrence rate can be improved by two to three digits as compared with that of FIG. As a result, a highly reliable E
An EPROM can be realized.

However, since the 1-bit data is stored by the two serial circuits, the storage capacity becomes 1/2 of the normal capacity. Therefore, it is difficult to increase the capacity. Further, the sense amplifier, peripheral circuits, and the like also become complicated.

[0024]

In the conventional device described above, each memory cell is configured to have the same read / write characteristics. Therefore, the reliability required for each data to be read / written is not the same. In other words, it is not uncommon for one piece of data not to be required to have such high reliability, but for another data to be required to have higher reliability. In order to meet such demands, conventionally, all the memory cells have been configured to be capable of high reliability. However, this is wasteful because data that does not require high reliability is read and written to the highly reliable memory cell.

In this case, the present invention has been made in consideration of the circumstances, and an object thereof is to make it possible to access data as a read / write object to a memory cell according to the reliability required thereof. Accordingly, it is to provide a highly reliable semiconductor memory device without waste.

[0026]

A first non-volatile semiconductor device of the present invention has a first memory region composed of a non-volatile memory cell having a control gate and a floating gate, and a control gate and a floating gate. A second memory area composed of a non-volatile memory cell and written by a data writing method such that the reliability of read / write characteristics is higher than that of the first memory area; and the first and second memory areas. Of the plurality of non-volatile memory cells configuring the same column, the plurality of data lines connected to the same column and the plurality of non-volatile memory cells configuring the first and second memory regions have the same row. A plurality of data lines connected to one another, a column decoder for selecting the data lines, and a row decoder for selecting the word lines It is.

The second non-volatile semiconductor device of the present invention comprises a plurality of non-volatile memory cells having a control gate and a floating gate, and a plurality of data to which the same column among the plurality of non-volatile memory cells is connected. Line, a plurality of word lines to which those of the plurality of nonvolatile memory cells having the same row are connected, a column decoder for selecting the data line, and a row decoder for selecting the word line, The plurality of non-volatile memory cells are divided into a first memory area and a second memory area in which writing is performed by a data writing method such that the read / write characteristics are more reliable than the first memory area. Configured as one.

A third non-volatile semiconductor device of the present invention is composed of a non-volatile memory cell having a control gate and a floating gate, and has a first reliability with respect to read / write characteristics. And a second non-volatile memory cell having a control gate and a floating gate, which is inferior to the first reliability in read / write characteristics.
A second memory region having high reliability, a plurality of data lines to which columns of the plurality of non-volatile memory cells having the same column are connected, and a row of the memory cells having the same row that are connected to each other. It is configured to include a plurality of word lines, a column decoder that selects the data lines, and a row decoder that selects the word lines.

[0029]

EXAMPLES First, the process until the examples of the present invention are obtained will be described.

21 (a) and 21 (b) show the EEP considered by the present inventor, in which each memory cell can be miniaturized.
2 shows a part (one cell, one-bit data storage) of a ROM. FIG. 21A is a plan view showing an actual arrangement state. Here, a portion surrounded by alternate long and short dash lines a, b, c, and d indicates one memory cell. FIG. 21B is a cross-sectional view taken along line AA of FIG. FIG. 21 differs from FIG. 15 in that the insulating oxide film 94 of the floating gate transistor 2 is a thin film having a thickness of about 100 A in FIG. 21 and the thin film portion 94A in FIG. In FIGS. 21 (a) and 21 (b), parts similar to those in FIG. 15 are denoted by the same reference numerals as in FIG.

With this structure, the size of the floating gate transistor 2 in the thickness direction can be significantly reduced. However, the lateral dimensions are
As can be seen from FIG. That is, the size of the cell is determined by the dimension l _{1 of} the contact 90, the margin l ₂ between the contact and the drain n ⁺ diffusion layer 90, and the dimension l ₃ of the isolation field portion between the adjacent drain n ⁺ diffusion layer. These dimensions are defined in a defined manufacturing process. Therefore, it is practically difficult to arbitrarily reduce the above dimensions. Therefore, if an attempt is made to configure the above-mentioned highly reliable nonvolatile semiconductor memory (1-bit data storage body) by using the two memory cells of FIG. 21, the chip will still be large and it will be difficult to increase the capacity. Conceivable.

The apparatus shown in FIG. 1 is based on FIG. FIG. 1A is different from FIG. 21A in that one contact 90 is commonly formed for two cells having a NAND configuration, that is, a cell having transistors 1A and 2A and a cell having transistors 1B and 2B. It is in the point which did. The equivalent circuit is shown in FIG. When the number of the contacts 90 is one, the lateral dimension of the memory cell is not determined by the contact portion 90 but is determined by the width w ₁ of the floating gate 95 and the distance w ₂ between the floating gates 95. That is, since the lateral dimension of the cell is determined by the minimum processing standard of the floating gate 95, the overall size of the device is significantly reduced. Comparing the sizes of the actual memory cells, the dashed lines a, b, c, and d in FIG. 1 are different from those in the case of using two memory cells shown in FIG.
The area of the memory cell (1-bit data storage) surrounded by is reduced to about 70%. FIG. 1B is a cross-sectional view taken along the line AA of FIG.

Next, the operation of the memory cell of FIG. 1 will be described.

In the equivalent circuit of FIG. 1C, erase and write operations are performed in the same manner as the memory cell of FIG. In the read mode, the word line WL is set at 5V, the data line DL is set at about 1V, and the control gate CG is set at 0V. It is assumed that data "1" is stored in a memory cell. That is, it is assumed that electrons are injected into the floating gates 95 of the two transistors 2A and 2B, respectively, and the threshold value of the transistors 2A and 2B is 10V. Then, no current flows through the two transistors 2A and 2B at the time of reading.

It is also assumed that data "0" is stored. That is, if holes are injected into the floating gates 95, 95 of the two transistors 2A, 2B, the threshold value of the floating gate transistors 2A, 2B will be, for example, -5V. This allows the two transistors 2
Both A and 2B are turned on, and current flows through the data lines DL to the transistors 1A and 2A; 1B and 2B.

Now, the oxide film 94 of the one transistor 2A
It is assumed that the holes in the floating gate 95 emit a negative charge due to a defect. At this time, if the stored data is "1", electrons are emitted from the floating gate 95, and the threshold value of the transistor 2A becomes 1V which is an initial value. However, since the control gate CG is at 0 V, the transistor 2A remains off. Therefore, no malfunction occurs. Conversely, if the stored data is "0",
Holes are emitted from the floating gate 95, the threshold value also changes from -5V to 1V, and the transistor 2A is turned off. However, the threshold value of the other normal transistor 2B is still -5V. As a result, a current flows through the data line DL via the transistor 2B, so that no malfunction occurs.

As described above, even if one of the two transistors, for example, the transistor 2A is defective, if the other transistor 2B is normal, the entire memory cell operates normally.

Although FIG. 2 shows the circuit configuration of the 1-bit type, it is also possible to configure the multi-bit type. FIG. 3 shows a 4-bit type, and shows a portion corresponding to a portion surrounded by a broken line in FIG. With this configuration, input / output of 4-bit data is performed.

4 and 5 show another example of the apparatus. FIG.
1A shows a portion (1 bit data storage body) corresponding to a portion surrounded by broken lines a, b, c and d in FIG.
4 and 5 are different from FIG. 1 in that the number of select gate transistors 1 is one. With this configuration, as shown in FIG. 4A, the area of the N-type diffusion layer region 93 connected to the data line DL can be reduced. As a result, the parasitic capacitance of the diffusion layer 93 is reduced, and the data line D
High-speed operation can be achieved by speeding up the charge and discharge of L. FIG.
The memory cell size in the case of (a) is the same as that of FIG. 1 (a). FIG. 5 is a cross-sectional view taken along the line AA of FIG. FIG. 4C is an equivalent circuit of FIG. 5, and FIG.
FIG. 3 is a circuit configuration diagram of an EEPROM.

FIG. 6 shows another example of the apparatus. This device example shows an example suitable for miniaturization. The example of FIG. 6 is shown in FIGS.
The difference from the example is that the two floating gate transistors 2
The point is that the sources A and 2B are separated into sources S _A and S _B. Source wiring of two sources S _A and S _B (A
1) has two lines in the layout of the drawing. However,
The sources S _A and S _B may be connected to the same source line.
With such a configuration of FIG. 3, the memory cell size is
Compared to the one shown in FIG. 21, it is 63%, which can be greatly reduced.

FIGS. 8A and 8B show still another example of the apparatus. In this example, the memory cell size can be further reduced by using the third polycrystalline silicon layer. That is, as can be seen particularly from FIG. 8B, the floating gate 95 is made of polycrystalline silicon of the first layer.
Then, the control gate 97 is formed of the second-layer polycrystalline silicon. Thereafter, an insulating film 102 is formed, and then a select gate 103 serving as a word line is formed of a third layer of polycrystalline silicon. With such a configuration, the diffusion layer 92 between the floating gate transistor 2 and the selection transistor 1 in FIG. 21 can be eliminated.
Since the diffusion layer 92 can be eliminated, the cell size can be further reduced. That is, the memory cell size can be 56% as compared with that of FIG. Further, FIG.
As can be seen from the above, the N layer 96A is provided between the floating gate 95 and the control gate 97, and the insulating film is O--N--O.
(Oxide-Nitride-Oxide) has a three-layer structure. With such a three-layer structure, two gates 95,
Even if the insulating film between 97 is thinned, the withstand voltage can be set high. The equivalent circuit of FIG. 8A is shown in FIG.

FIG. 7 shows a modification of FIG. 8 in the same cross section as FIG. 4 (b). In FIG. 8B, the floating gate 95
Is etched using the gate 97 as a mask following the formation of the control gate 97 when the control gate 97 is formed, so that the gate 97 is formed to have substantially the same dimensions. Thereafter, when the third-layer selection gate 103 is formed, the selection gate 103 and the floating gate 95 are directly opposed to each other, and in some cases, the breakdown voltage between the floating gate 95 and the selection gate 103 may be deteriorated. Is concerned. FIG. 7 shows an improvement. First, the floating gate 95 is formed, and then the control gate 97 is formed so as to sufficiently cover the floating gate 95. In the figure, reference numerals 91A and 92A denote N-type diffusion layers, which may have a slightly lower concentration than the diffusion layers 91 and 93. With such a configuration, the floating gate 95 is completely covered with the ONO insulating film. Thereby, the breakdown voltage between the floating gate 95 and the selection gate 103 is improved, and the reliability is also improved.

The defective mode described above is a mode in which the charge in the floating gate disappears due to the current leakage caused by the deterioration of the insulating oxide film and the like between the floating gate 95 and the semiconductor substrate 80 and the defect. However, if the deterioration is severe, the W / E (write / read) may be repeated to cause complete destruction. At the time of this complete breakdown, the floating gate 95 and the drain 92 are completely short-circuited. Thus, the potential of the floating gate 95 becomes equal to the potential of the drain 92 irrespective of the potential of the control gate 97. Even if such a defect occurs, the effect of the present invention remains unchanged if the voltage of the drain 92 at the time of reading is set to 1 V or less. That is, the initial threshold value of the destroyed cell is 1V. Therefore, if the drain voltage at the time of reading is set to 1 V or less, the destroyed cell is always turned off at the time of reading. For this reason, the effect of the memory cell of the present invention can be exhibited.

FIG. 9 shows an example of a device (EEP) in which a 1-bit data storage body can be constituted by one transistor and which is suitable for miniaturization.
ROM). 9 is the same as that in FIG. 21 except that the selection transistor 1 is omitted, and has only a floating gate transistor 2 as a transistor. FIG. 9B is a sectional view taken along line AA of FIG. 9A, and FIG. 9C is an equivalent circuit of FIG. 9A.

Next, these operations will be described.

At the time of writing, a high voltage (for example, 7V) is applied to the drain D, 0V is applied to the source S, and a high voltage (for example, 12V) is applied to the control gate CG. Thereby, electrons are generated by the hot electron effect. Those electrons are injected into the floating gate. As a result, the threshold value of this transistor shifts in the positive direction, for example, to 8V. At the time of erasing, the drain D is floated and the control gate CG
At a low potential (for example, 0 V) and the source S at a high voltage (for example, 1 V).
2V) is applied. In this way, electrons in the floating gate are emitted to the source S due to the Fowler-Nordheim tunnel effect. This causes the threshold of this transistor to shift in the negative direction. In this case, if the data is erased too much, the threshold value becomes negative. For this reason, it is necessary to stop erasing at an appropriate point. Normally, the threshold value after erasing is set between 0 and 5V. Preferably, it is in the order of 1-2V. Usually, in this type of memory, a plurality of memory cells are collectively erased because the sources are commonly connected.

At the time of reading, about 1 V is applied to the drain D, 0 V is applied to the source S, and 5 V is applied to the control gate CG. At this time, if the cell is in the written state, this transistor is turned off and no current flows. On the other hand, if the cell is in the erased state, the cell turns on and a current flows. This is sensed by the sense amplifier to read the stored data.

Although such a memory cell is suitable for miniaturization, at the time of erasing, a plurality of memory cells (in some cases, all memory cells of the chip) are erased at once, and the threshold value is set to a constant value. Need to control. However, when a tunnel current flows through the oxide film during erasing, electrons are trapped in defects and the like in the oxide film, and writing and erasing (W / E) are repeated, thereby causing a defect that erasing characteristics deteriorate. come. Such defects often occur accidentally with a certain probability. For example, at the initial stage when W / E is performed about 10,000 times, in the case of a 1-Mbit memory, erasure failure occurs in about one to several bits.

FIG. 10 shows the entire EEPROM in which the erase failure is improved by using the cells of FIGS. 9 (a) to 9 (c).
In this device example, each 1 bit is composed of two memory cells 30A and 30B, as can be seen from the broken line 40. In this way, even if one memory cell accidentally causes an erasure failure, the other memory cell is normally erased. Therefore, all the memory cells are uniformly erased even when the entire chip is erased collectively. In the example of FIG. 10, the common source VS * is provided common to all cells.
However, the memory cell array may be divided into a plurality of blocks, and a common source may be provided in each of the blocks to perform erasing for each block.

FIG. 11 shows a plan view pattern diagram as a specific example of FIG. 10, and corresponds to a portion 7b of FIG. FIG. 1
In FIG. 1 and FIG. 10, the same members are designated by the same reference numerals. Further, abcd in FIG. 11 corresponds to abcd in FIG. 9A.

FIG. 12 shows an embodiment of the present invention which is a modification of FIG.

In this embodiment of the present invention, the memory cell array is divided into a first portion connected to the word lines WL1 to WLk and a second portion connected to the word lines WL (k + 1) to WLm. A first row decoder 32-1 for selecting the word lines WL1 to WLk and a word line WL (k + 1)
Second row decoder 32-2 for selecting .about.WLm is separately provided. And, in the first part, as shown in FIG.
Similar to the example of 11, the 1-bit data storage is composed of two cells to form a highly reliable memory area. The second part is a normal memory area in which a one-bit data storage is constituted by one cell. Such a configuration is particularly suitable for W / E
It can be said that the 2-cell / bit configuration is applied only to the region requiring high reliability. Therefore, it is possible to minimize the increase in the chip area while improving the reliability.

In this example, the common sources are VS * 1 and VS.
* 2, but these may be shared. Further, in this example, the data line is common. However, when the memory cells shown in FIGS. 1 to 8 are used as the memory cells, the lateral pitches of the memory cells of the first portion and the second portion are different from each other. For this reason,
The array may be completely separated in the first and second parts and each may be provided with a row decoder and a column decoder separately.

FIG. 13 shows another example of the apparatus. FIG.
In 3, the memory cells arranged in the row direction are all connected to the left and right, and a pair of left and right memory cells are set as a 1-bit data storage body as shown by a broken line 40.

That is, in FIG. 13, memory cells arranged in the row direction are sequentially connected in series. That is, the drain D of a certain memory cell 30-1 and the memory cell 30-
2 is connected to the source S of a certain memory cell 30-1 and the source S of a memory cell 30-3 on the right side thereof. In other words, for two adjacent memory cells, the drain of one cell and the drain of another cell are connected to each other, and for the two other adjacent memory cells, the source of one cell and the source of another cell Are connected to each other. The data lines DL1 to DLn are connected to the drain D of each memory cell, and the common sources S * 1 to S * (n + 1) are connected to the source S. These common sources S * 1 to S * (n + 1) are further connected to a common source VS *. As a result, a pair of left and right memory cells 30-1 and 30-2 constitute a 1-bit memory cell, as illustrated by being surrounded by a broken line 40 in FIG.

An example of the actual layout of the embodiment shown in FIG. 13 is shown in FIGS. In particular, as can be seen from FIG. 3B, the source S and the drain D are arranged in the word line direction.
Diffusing layers are alternately provided. These diffusion layers are shared for two adjacent transistors. That is,
For example, focusing on the memory cells 30-1 and 30-2, the drain D1 existing between them is shared as the drains D and D of the two memory cells 30-1 and 30-2. The source S1 existing between the memory cells 30-1 and 30-3 is shared as the sources S and S of these two memory cells. That is, a field oxide film for isolation is not required between each memory cell, and does not exist at present. Therefore, miniaturization in the word line direction is achieved.

In particular, as can be seen from FIG. 14A, the data lines DL1, DL2, ... And the common source (source wiring) S * 1, S * 2 ,. There is. These data lines and source lines are connected to the diffusion layers (source, drain) at predetermined intervals by contacts 90, 90,.... The distance between the contacts is set so that the resistance of the diffusion layer of the drain or the source does not affect the characteristics.

Most of the embodiments described above are of the 1-bit type. However, FIG.
As a matter of course, it can be of a multi-bit type as shown in (e).

Although FIGS. 1 to 6 show an example in which the gates of the cell selection transistors 1, 1A and 1B are formed of the second conductive layer (for example, polysilicon), a floating gate is formed, for example. The first conductive layer (for example, polysilicon) and the second conductive layer are formed into a two-layer structure.
The insulating film between the second conductive layer and the second conductive layer may be etched and short-circuited. By doing so, the selection transistors 1, 1A, and 1B can be formed in the same process as that for forming the floating gate transistor 2, so that the processing margin is improved.

[0060]

According to the present invention, since the first and second memory areas having different read / write characteristics are provided, the read / write reliability required for the data to be accessed is improved. Since it is possible to access the corresponding region, it is possible to provide a highly reliable semiconductor memory device with no waste.

[Brief description of drawings]

FIG. 1 is a partial plan view of an example of an apparatus related to the present invention, and a sectional view taken along the line AA.

FIG. 2 is an equivalent circuit diagram thereof.

FIG. 3 is an overall circuit diagram thereof.

FIG. 4 is a partial plan view of another device example, a cross-sectional view taken along the line AA, and its equivalent circuit diagram.

FIG. 5 is an overall circuit diagram thereof.

FIG. 6 is a partial plan view of another device example and its equivalent circuit diagram.

FIG. 7 is a partial cross-sectional view of another example of the apparatus.

FIG. 8 is a partial plan view of another device example, AA.
A line sectional view and its equivalent circuit diagram.

FIG. 9 is a partial plan pattern diagram of another example of the apparatus, and its AA
A line sectional view and its equivalent circuit diagram.

FIG. 10 is an overall circuit diagram of another device example.

FIG. 11 is a partial plan view thereof.

FIG. 12 is an overall circuit diagram of an embodiment of the present invention.

FIG. 13 is an overall circuit diagram of another device example.

FIG. 14 is a plan pattern diagram, a sectional view taken along the line AA and a sectional view taken along the line BB of a part of the actual device configured based on FIG.

FIG. 15 is a cross-sectional view of a conventional memory cell.

FIG. 16 is an equivalent circuit diagram thereof.

FIG. 17 is a characteristic diagram thereof.

FIG. 18 is a characteristic diagram thereof.

FIG. 19 is an overall circuit diagram of a conventional device.

FIG. 20 is an overall circuit diagram of another example of the conventional device.

FIG. 21 is a plan view of a memory cell according to the creation of the present inventor and a cross-sectional view taken along the line AA.

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁶ Identification code Agency reference number FI Technical display location H01L 29/792

Claims (3)

[Claims] 1. A first memory region comprising a non-volatile memory cell having a control gate and a floating gate, and a non-volatile memory cell comprising a control gate and a floating gate, wherein the reliability of read / write characteristics is the same. A second memory area that is written by a data writing method that is higher than the first memory area, and a plurality of nonvolatile memory cells that form the first and second memory areas and have the same column A plurality of data lines connected to each other, a plurality of data lines connected to the same row among the plurality of nonvolatile memory cells forming the first and second memory regions, and selecting the data line And a row decoder for selecting the word line. 2. A plurality of non-volatile memory cells having a control gate and a floating gate, a plurality of data lines to which the same column among the plurality of non-volatile memory cells is connected, and the plurality of non-volatile memory cells. A plurality of word lines to which those having the same row are connected, a column decoder that selects the data line, and a row decoder that selects the word line, and the plurality of nonvolatile memory cells are A non-volatile semiconductor memory characterized by being divided into a memory area and a second memory area in which writing is performed by a data writing method such that the read / write characteristics are more reliable than the first memory area. apparatus. 3. A non-volatile memory cell including a non-volatile memory cell having a control gate and a floating gate, having a first reliability for read / write characteristics, and a non-volatile memory having a control gate and a floating gate. Memory cell and has a read / write characteristic of
A second memory region having a second reliability lower than that of a plurality of data lines, a plurality of data lines to which the same column of the plurality of nonvolatile memory cells is connected, and a row of the plurality of memory cells. A non-volatile semiconductor memory device comprising: a plurality of word lines connected to each other, a column decoder selecting the data line, and a row decoder selecting the word line.