Non-volatile semiconductor memory device with a memory array preventing generation of a through current path

1. Field of the Invention

The present invention generally relates to a non-volatile semiconductor memory device. More particularly, the present invention relates to the memory array structure of a flash memory.

2. Description of the Background Art

Recently, a flash memory capable of non-volatile data storage (batch-erasable, electrically rewritable read only memory) goes mainstream. In particular, a MONOS (Metal Oxide Nitride Oxide Silicon) memory transistor cell has attracted attention in such a flash memory because of its low costs, small area, and charge storage portion. The MONOS memory transistor cell is characterized in that a nitride film is used as the charge storage portion instead of a floating gate used in the conventional examples.

The MONOS memory transistor cell is different from the floating gate: structure in that the polycrystalline silicon of floating gate is replaced with a nitride film.

FIGS. 10A, 10 B and 10 C illustrate normal write, read and erase operations of the MONOS memory transistor cell, respectively.

Referring to FIGS. 10A, 10 B and 10 C, MONOS memory transistor cell MC is formed from a stacked layer of three insulating films (i.e., an oxide film; 4 , a nitride film 5 , and an oxide film 6 ) and a polycrystalline silicon of control gate 7 on a P-type semiconductor substrate 1 . Near the stacked gate on P-type semiconductor substrate 1 , N-channel diffusion regions 2 , 3 are formed in a self-aligned manner. Memory transistor cell MC thus corresponds to an N-channel field effect transistor formed on P-type semiconductor substrate 1 .

FIG. 10A illustrates normal write operation of MONOS memory transistor cell MC.

When 0 V is applied to P-type semiconductor substrate 1, 10 V is applied to control gate 7 , 5 V is applied to diffusion region 2 , and 0 V is applied to diffusion region 3 , channel electrons are accelerated by a steep electric field created in diffusion region 2 of the memory transistor cell. Those accelerated electrons which overcome the barrier height of the oxide film are trapped on the side of diffusion region 2 in nitride film 5 (bit 1 ). As a result, the threshold voltage of memory transistor cell MC is increased to render memory transistor cell MC in the written state storing data “0”. For example, the data of bit 1 is “0” in the written state caused by the trapped electrons, and “1” in the state where no data is written (i.e., in the erased state). The following description is given on the assumption that the data is “0” in the written state and “1” in the erased state.

When the applied voltages to diffusion regions 2 , 3 are reversed, i.e., when 0 V is applied to diffusion region 2 and 5 V is applied to diffusion region 3 as shown in parentheses in FIG. 10A, the electrons are trapped on the side of diffusion region 3 in nitride film 5 (bit 2 ). As a result, the threshold voltage of memory transistor cell MC is increased to render memory transistor cell MC in the written state storing data “0”. Accordingly, the data of bit 2 is “0” in the written state caused by the trapped electrons, and “1” in the erased state.

This MONOS structure traps electrons in non-covalent bonds (dangling bonds) that are distributed in nitride film 5 at random. Electrons are stored at different positions in nitride film 5 (i.e., on the sides of diffusion regions 2 and 3 in nitride film 5 ) one bit each, enabling two-bit/cell data storage.

FIG. 10B illustrates read operation of MONOS memory transistor cell MC.

First, operation of reading bit 1 on the side of diffusion region 2 in nitride film 5 will be described.

A voltage of 0 V is applied to P-type semiconductor substrate 1 , 3 V is applied to control gate 7 , 0 V is applied to diffusion region 2 , and 2 V is applied to diffusion region 3 . It is herein assumed that memory transistor cell MC is in the written state on the side of diffusion region 2 in nitride film 5 , that is, electrons have been trapped in nitride film 5 . In this case, a high threshold voltage inhibits memory transistor cell MC from being turned ON, and no current path is formed from diffusion region 3 to diffusion region 2 . As a result, “0” can be read as the data of bit 1 . On the other hand, when memory transistor cell MC is in the erased state on the side of diffusion region 2 in nitride film 5 , a low threshold voltage allows memory transistor cell MC to be turned ON, and a current path is formed from diffusion region 3 to diffusion region 2 . As a result, “1” can be read as the data of bit 1 .

Hereinafter, operation of reading the data of bit 2 on the side of diffusion region 3 in nitride film 5 will be described.

The applied voltages to diffusion regions 2 , 3 are reversed. More specifically, 0 V is applied to P-type semiconductor substrate 1 , 3 V is applied to control gate 7 , 2 V is applied to diffusion region 2 , and 0 V is applied to diffusion region 3 , as shown in parentheses in FIG. 10 B. It is herein assumed that memory transistor cell MC is in the written state on the side of diffusion region 3 in nitride film 5 , that is, electrons have been trapped in nitride film 5 . In this case, a high threshold voltage inhibits memory transistor cell MC from being turned ON, and no current path is formed from diffusion region 2 to diffusion region 3 . As a result, “0” can be read as the data of bit 2 . On the other hand, when memory transistor cell MC is in the erased state on the side of diffusion region 3 in nitride film 5 , a low threshold voltage allows memory transistor MC to be turned ON, and a current path is formed from diffusion region 2 to diffusion region 3 . As a result, “1” can be read as the data of bit 2 .

Accordingly, by adjusting the applied voltages to diffusion regions 2 , 3 , bit 1 and bit 2 can be read according to whether a current path is formed or not. This enables two-bit/cell read operation.

FIG. 10C illustrates erase operation of MONOS memory transistor cell MC.

First, operation of erasing bit 1 on the side of diffusion region 2 in nitride film 5 will be described.

It is herein assumed that 0 V is applied to P-type semiconductor substrate 1 and control gate 7 , 10 V is applied to diffusion region 2 , and diffusion region 3 is in the open state (OPEN).

In this case, it flows a Fowler-Nordheim current from bit 1 on the side of diffusion region 2 trapped the electrons in nitride film 5 to substrate region 1 or diffusion region 2 . The electrons are thus removed from the side of diffusion region 2 in nitride film 5 . In this state, memory transistor cell MC has a reduced threshold voltage.

Hereinafter, operation of erasing bit 2 on the side of diffusion region 3 in nitride film 5 will be described.

It is herein assumed that 0 V is applied to P-type semiconductor substrate 1 and control gate 7 , 10 V is applied to diffusion region 3 , and diffusion region 2 is in the open state, as shown in parentheses in FIG. 10 C.

In this case, it flows a Fowler-Nordheim current from bit 2 on the side of diffusion region 3 trapped the electrons in nitride film 5 to substrate region 1 or diffusion region 3 . The electrons are thus removed from the side of diffusion region 3 in nitride film 5 . In this state, memory transistor cell MC has a reduced threshold voltage.

Note that, when 10 V is applied to both diffusion regions 2 , 3 , electrons are removed from both bits 1 , 2 . The data is erased in this way.

FIG. 11 shows the structure of an array of a flash memory using the above MONOS memory transistor cells MC (hereinafter, referred to as NROM(R) memory array).

Referring to FIG. 11, the NROM(R) memory array includes identical memory block units MBU arranged in a matrix. Adjacent two memory block units MBU are electrically coupled to each other.

Memory block unit MBU 1 located in the center will now be described. Since the other memory block units MBU have the same circuit structure as that of memory block unit MBU 1 , detailed description thereof will not be repeated.

Memory block unit MBU 1 includes memory transistor cells MC( 0 . 0 ) to MC( 31 . 3 ) arranged in a matrix. Memory transistor cell MC(i.j) herein indicates a memory transistor cell located at the intersection of the i th row and the j th column (where i, j is an integer equal to or larger than zero).

Memory block unit MBU 1 further includes word lines WL 0 to WL 31 (hereinafter, generally referred to as word lines WL), diffusion bit lines BL 0 to BL 3 (hereinafter, generally referred to as diffusion bit lines BL), main bit lines MBL 0 , MBL 1 (hereinafter, generally referred to as main bit lines MBL), and gate units gtu 1 , gtu 2 . Word lines WL 0 to WL 31 correspond to the rows of memory transistor cells MC( 0 . 0 ) to MC( 31 . 3 ), respectively. Diffusion bit lines BL 0 to BL 3 correspond to the columns of memory transistor cells MC( 0 . 0 ) to MC( 31 . 3 ), respectively. Gate unit gtu 1 electrically couples diffusion bit lines BL 0 , BL 2 to main bit line MBL 0 . Gate unit gtu 2 electrically couples diffusion bit lines BL 1 , BL 3 to main bit line MBL 1 .

Gate unit gtu 1 includes a gate transistor gt 1 for electrically coupling diffusion bit line BL 0 to main bit line MBL 0 , and a gate transistor gt 2 for electrically coupling diffusion bit line BL 2 to main bit line MBL 0 . Gate transistors gt 1 , gt 2 are activated by gate select lines gs 1 , gs 2 , respectively.

Gate unit gtu 2 includes a gate transistor gt 3 for electrically coupling diffusion bit line BL 1 to main bit line MBL 1 , and a gate transistor gt 4 for electrically coupling diffusion bit line BL 3 to main bit line MBL 1 . Gate transistors gt 3 , gt 4 are activated by gate select lines gs 3 , gs 4 , respectively.

A plurality of non-volatile memory transistor cells MC arranged in the same column are herein collectively referred to as a column memory unit MU. For example, memory transistor cells MC( 0 . 0 ) to MC( 31 . 0 ) forms a column memory unit MU.

Each column memory unit MU is arranged between adjacent two of diffusion bit lines BL 0 to BL 4 . Adjacent two column memory units MU share diffusion bit line BL located therebetween. Column memory unit MU of memory transistor cells MC( 0 . 0 ) to MC ( 31 . 0 ) shares diffusion bit line BL 0 with column memory unit MC in the left memory block unit MBU.

Memory block unit MBU 2 includes memory transistor cells MC( 0 . 4 ) to MC( 31 . 7 ) arranged in a matrix, gate units gtu 3 , gtu 4 , diffusion bit lines BL 4 to BL 7 , and main bit lines MBL 2 , MBL 3 . Gate unit gtu 3 includes gate transistors gt 5 , gt 6 . Gate unit gtu 4 includes gate transistors gt 7 , gt 8 . Since the connection between the above elements is the same as that in memory block unit MBU 1 , detailed description thereof will not be repeated. Column memory unit MU of memory transistor cells MC( 0 . 7 ) to MC( 31 . 7 ) is electrically coupled to diffusion bit line BL included in the right memory block unit MBU. Note that memory block unit MBU generally refers to memory block units MBU 1 , MBU 2 and the like.

The above NROM(R) memory array structure allows two-bit/cell read operation by using small MONOS memory transistor cells, shares a bit line between adjacent two column memory units MU, and uses a diffusion layer for the bit lines. This enables improvement in integration.

Hereinafter, operation of simultaneously reading bit 1 on the side of diffusion region 2 of memory transistor cell MC( 0 . 1 ) of memory block unit MBU 1 (hereinafter, generally referred to as right bit) and the right bit of memory transistor cell MC( 0 . 5 ) of memory block unit MBU 2 will be considered.

FIG. 12 is a fining chart of the operation of simultaneously reading the respective right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 5 ).

Referring to FIG. 12, in the operation of reading the right bits, main bit lines MBL 1 , MBL 3 first rise to 2 V. The other main bit lines MBL 0 , MBL 2 are at 0 V. Gate select lines gs 2 , gs 3 then rise to 8 V (“H” level). In response to this, gate transistors gt 3 , gt 7 are turned ON, and the voltage (2 V) on main bit lines MBL 1 , MBL 3 is transmitted to diffusion bit lines BL 1 , BL 5 . Gate transistors gt 2 , gt 6 are also turned ON, and diffusion bit lines BL 2 , BL 6 are electrically coupled to main bit lines MBL 0 , MBL 2 , respectively. The other word lines WLs other than word line WL 0 are at 0 V. In the operation of reading the right bits, gate select lines gs 1 , gs 4 are at 0 V.

At time t 1 , word line WL 0 is activated and rises to 3 V. The right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 5 ) are read in this way.

It is herein assumed that the right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 5 ) are in the erased state. In this case, the right-bit read operation is performed as follows:

Memory transistor cells MC( 0 . 1 ), MC( 0 . 5 ) are turned ON in response to the voltage from word line WL 0 , because of their low threshold voltage. This raises the voltage levels of diffusion bit lines BL 2 , BL 6 and thus the voltages of main bit lines MBL 0 , MBL 2 electrically connected thereto, respectively. Once the voltages rise to such a level that allows a not-shown sense amplifier connected to main bit lines MBL 0 , MBL 2 to read the data, that is, at time t 2 , the voltages on main bit lines MBL 1 , MBL 3 are caused to fall.

The voltages of diffusion bit lines BL 1 , BL 5 fall responsively. At time t 3 the voltages of main bit lines MBL 1 , MBL 3 and diffusion bit lines BL 1 , BL 5 fall to 0 V, the voltages of gate select lines gs 2 , gs 3 and word line WL 0 are caused to fall.

It is herein assumed that the right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 5 ) have already been in the written state. In this case, the right-bit read operation is performed as follows:

Memory transistor cells MC( 0 . 1 ), MC( 0 . 5 ) will not be turned ON in response to the rise of word line WL 0 at time t 1 , because of their high threshold voltage.

However, if MC( 0 . 2 ) to MC( 0 . 4 ) of memory transistor cells MC( 0 . 1 ) to MC(O. 5 ) are in the erased state, all memory transistor cells MC( 0 . 2 ) to MC( 0 . 4 ) are turned ON because of their low threshold voltage.

The voltages of diffusion bit lines BL 5 , BL 2 are 2V and 0 V, respectively. Adjacent diffusion bit lines BL are electrically connected to each other through corresponding memory transistor cells MC. Therefore, a current path is formed from diffusion bit line BL 5 to diffusion bit line BL 2 in response to the rise of word line WL 0 . In other words, a through current path is formed. Provided that the right bit of memory transistor cell MC( 0 . 1 ) is in the written state, a through current path is formed from diffusion bit line BL 5 to diffusion bit line BL 2 even if memory transistor cell MC( 0 . 1 ) is not turned ON. This raises the voltage levels of diffusion bit line BL 2 and main bit line MBL 0 . As a result, the sense amplifier connected to main bit line MBL 0 may erroneously recognize the written state of the right bit of memory transistor cell MC( 0 . 1 ) as the erased state.

In the operation of simultaneously writing data to memory transistor cells MC( 0 . 1 ), MC( 0 . 5 ), a similar through current path may possibly be formed. Such a through current path may cause the data to be erroneously written to an unintended memory transistor cell MC.

It is an object of the present invention to provide a non-volatile semiconductor memory device having a memory array structure that prevents generation of a through current path.

A non-volatile semiconductor memory device according to the present invention includes a memory array including a plurality of non-volatile memory transistor cells arranged in a matrix. The plurality of non-volatile memory transistor cells each has a threshold voltage according to a storage data level. The memory array is divided into a plurality of memory blocks. Each of the plurality of memory blocks includes a plurality of memory cell columns, and a plurality of bit lines arranged along column direction on both sides of each corresponding memory cell column, respectively. Each of the bit lines is arranged between adjacent two memory cell columns being shared by the adjacent two memory cell columns. The memory array includes a plurality of isolating portions each provided between adjacent two memory blocks, for electrically isolating two memory blocks from each other.

Preferably, a read operation and a write operation are performed one bit per memory block.

Accordingly, a main advantage of the present invention is as follows: each isolating portion provided between adjacent memory blocks electrically isolates the corresponding memory blocks from each other. Therefore, when write operation and read operation are performed one bit per memory block, no through current path (current path between memory blocks) is formed. As a result, erroneous write and read operation can be prevented.

The foregoing and other objects, features, aspects and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings.

FIG. 1 shows the overall structure of a non-volatile semiconductor memory device 1000 according to the present invention.

FIG. 2 shows the structure of an NROM(R) memory array 1001 according to a first embodiment of the present invention.

FIG. 3 is a timing chart illustrating operation of reading the right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 4 ).

FIG. 4 is a timing chart illustrating operation of writing to the right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 4 ).

FIG. 5 is a timing chart illustrating erase operation in the first embodiment.

FIG. 6 shows the structure of another structural example of memory array 1001 , that is, an NROM(R) memory array 1100 according to a second embodiment of the present invention.

FIG. 7 is a flowchart illustrating an example of a common erase sequence of a flash memory.

FIG. 8 shows the structure of still another structural example of memory array 1001 , that is, an NROM(R) memory array 1200 according to a third embodiment of the present invention.

FIG. 9 shows the overall structure of a non-volatile semiconductor memory device 2000 according to a fourth embodiment of the present invention.

FIGS. 10A, 10 B and 10 C illustrate write, read and erase operations of a MONOS memory transistor cell MC, respectively.

FIG. 11 shows the array structure of a flash memory using the MONOS memory transistor cells MC.

FIG. 12 is a timing chart illustrating operation of simultaneously reading the right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 5 ).

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. Note that the same or corresponding portions are denoted with the same reference numerals and characters throughout the figures, and description thereof will not be repeated.

First Embodiment

Referring to FIG. 1, a non-volatile semiconductor memory device 1000 includes a memory array 1001 , an X decoder 1002 , a Y decoder 1005 , a Y gate 1004 , and a data register 1003 . Memory array 1001 includes a plurality of non-volatile memory transistor cells MC arranged in a matrix. X decoder 1002 selects a row of memory array 1001 according to an X address. Y decoder 1005 generates a column select signal for selecting a column of memory array 1001 . Y gate 1004 connects the selected column of memory array 1001 to an internal data line according to the column select signal from Y decoder 1005 . Data register 1003 temporarily stores write data in write operation. Data register 1003 includes register circuits corresponding to the respective columns, and stores write data received through Y gate 1004 . The write operation is performed based on the data stored in data register 1003 .

Non-volatile semiconductor memory device 1000 further includes a data input/output (I/O) buffer 1007 , an address input buffer 1008 , and a control signal input buffer circuit 1010 . Data I/O buffer 1007 receives and outputs data from and to the outside through a data/address terminal group 1006 . Address input buffer 1008 generates an internal address signal (X and Y address signals) in response to an address signal received through data/address terminal group 1006 . Control signal input buffer circuit 1010 generates an internal control signal in response to a control signal received through a control terminal group 1009 .

The control signals applied to control terminal group 1009 include an output enable signal OE giving an instruction to perform read operation, a chip enable signal CE indicating that non-volatile semiconductor memory device 1000 is selected, and a write enable signal WE designating the timing of writing the data or fetching a command.

In response to activation of chip enable signal CE, control signal input buffer circuit 1010 determines that the external control signals are valid, and fetches data/address. In response to output enable signal OE and chip enable signal CE, control signal input buffer circuit 1010 outputs an internal output enable signal and an internal chip enable signal to data I/O buffer 1007 and address input buffer 1008 . The read operation is performed in response to activation of both output enable signal OE and chip enable signal CE.

Semiconductor memory device 1000 further includes an X address buffer 1011 , a data input/output (I/O) control circuit 1012 , a command control circuit 1013 , a read/write/erase control circuit 1014 , and a voltage generator circuit 1015 . X address buffer 1011 buffers an internal address signal from address input buffer 1008 , and generates an internal X address signal for output to X decoder 1002 . In write operation, data I/O control circuit 1012 buffers write data from data I/O buffer 1007 for writing to data register 1003 through Y gate 1004 . In read operation, data I/O control circuit 1012 amplifies internal read data received through Y gate 1004 for output to data I/O buffer 1007 . Command control circuit 1013 determines a designated operation mode in response to an internal control signal from control signal input buffer circuit 1010 and a command from address input buffer 1008 . Read/write/erase control circuit 1014 generates a control signal required for the designated operation mode, according to an instruction from command control circuit 1013 . Voltage generator circuit 1015 generates a voltage required for read/write/erase operation under the control of read/write/erase control circuit 1014 .

For example, when the internal chip enable signal is active, command control circuit 1013 fetches a command from address input buffer 1008 in response to the rise of an internal write enable signal, and decodes the command. The internal write enable signal is generated by control signal input buffer circuit 1010 in response to write enable signal WE. Read/write/erase control circuit 1014 , which is formed from a processor such as a sequence controller, generates a control signal required for operation designated by command control circuit 1013 in order to control operation of each circuit.

Voltage generator circuit 1015 generates a voltage required for read/write/erase operation. A high voltage generated by voltage generator circuit 1015 may be applied to different portions of the memory array. A voltage is actually applied to the memory transistor cells and the substrate region through X decoder 1002 and Y decoder 1005 .

Read/write/erase control circuit 1014 controls acceptance of a command by command control circuit 1013 in read/write/erase operation. Read/write/erase control circuit 1014 controls the operation of X address buffer 1011 in order to sequentially change X address to select the memory transistor cells when verifying batch erase operation.

The non-volatile memory transistor cells arranged in a matrix in memory array 1001 are MONOS memory transistor cells. Information is stored according to fluctuation in threshold voltage of the MONOS memory transistor cells.

Referring to FIG. 2, NROM(R) memory array 1011 is divided into a plurality of memory blocks MB arranged in a matrix. Write and read operations are performed one memory transistor cell MC per memory block MB.

Hereinafter, memory block MB 1 located in the center will now be described.

Memory block MB 1 includes memory transistor cells MC( 0 . 0 ) to MC( 31 . 2 ) arranged in a matrix, diffusion bit lines BL 0 to BL 3 arranged in columns, word lines WL 0 to WL 31 arranged in rows, main bit lines MBL 0 , MBL 1 arranged in columns, and gate units gtu 1 , gtu 2 . Gate unit gtu 1 electrically couples diffusion bit lines BL 0 , BL 2 to main bit line MBL 0 . Gate unit gtu 2 electrically couples diffusion bit lines BL 1 , BL 3 to main bit line MBL 1 . Gate unit gtu 1 includes gate transistors gt 1 , gt 2 that are activated by gate select lines gs 1 , gs 2 , respectively. Gate unit gtu 2 includes gate transistors gt 3 , gt 4 that are activated by gate select lines gs 3 , gs 4 , respectively.

Memory block MB 1 of the first embodiment is different from memory block unit MBU 1 of the conventional example in that memory block MB 1 does not include column memory unit MU located on the right of diffusion bit line BL 3 , and in that diffusion bit lines BL 0 , BL 3 located at both ends are not shared diffusion bit lines.

More specifically, in the conventional example, adjacent two memory block units MBU are continuous in the boundary region therebetween, and diffusion bit lines BL arranged in columns are electrically coupled to each other through corresponding memory transistor cells MC. In the present embodiment, the boundary region between adjacent two memory blocks MB is a, completely electrically isolated region. For example, the isolated region can be formed from a means capable of electrical insulation such as an insulating film. Since the connection between the elements is the same as that described for memory block unit MBU 1 , detailed description thereof will not be repeated.

Memory block MB 2 includes memory transistor cells MC( 0 . 3 ) to MC( 31 . 5 ) arranged in a matrix, diffusion bit lines BL 4 to BL 7 , word lines WL 0 to WL 31 , main bit lines MBL 2 , MBL 3 , and gate units gtu 3 , gtu 4 . Gate unit gtu 3 includes gate transistors gt 5 , gt 6 , and gate unit gtu 4 includes gate transistors gt 7 , gt 6 . The connection between the above elements is the same as that of memory block MB 1 .

Hereinafter, operation of simultaneously reading the respective right bits of memory transistor cell MC( 0 . 1 ) of memory block MB 1 and memory transistor cell MC( 0 . 4 ) of memory block MB 2 will be described.

Referring to FIG. 3, in the operation of reading the right bits, main bit lines MBL 1 , MBL 3 first rise to 2 V, as described above in connection with FIG. 12 . Gate select lines gs 2 , gs 3 then rise to 8 V (“H” level). In response to this, gate transistors gt 3 , gt 7 are turned ON, and the voltage (2 V) on main bit lines MBL 1 , MBL 3 is transmitted to diffusion bit lines BL 1 , BL 5 . At time t 4 , word line WL 0 is activated, and rises to 3 V. As a result, the right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 4 ) are read. When the right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 4 ) are in the erased state, that is, when the right bits are “1”, memory transistor cells MC( 0 . 1 ), MC( 0 . 4 ) are turned ON. This raises the voltage levels of main bit lines MBL 0 , MBL 2 , whereby data “1” can be read. After a prescribed time, i.e., at time t 5 , the voltages of main bit lines MBL 1 , MBL 3 are caused to fall. In response to this, the voltages of diffusion bit lines BL 1 , BL 5 also fall. Once the voltages of diffusion bit lines BL 1 , BL 5 falls sufficiently at time t 6 , the voltages of gate select lines gs 2 , gs 3 and word line WL 0 are caused to fall.

When the right bits of memory transistor cells MC( 0 . 1 ), MC( 0 . 4 ) are in the written state, that is, when the right bits are “0”, a high threshold voltage hinders memory transistor cells MC( 0 . 1 ), MC( 0 . 4 ) from being turned ON in response to the rise of word line WL 0 at time t 4 .

Since adjacent two memory blocks MB are completely isolated from each other by the boundary region, a through path current path as in the conventional example will not be generated between memory blocks MB 1 and MB 2 , enabling correct read operation.

Accordingly, the NROM(R) memory array structure of the first embodiment enables erroneous read operation to be avoided.

Hereinafter, operation of simultaneously writing the data to the respective right bits of memory transistor cell MC( 0 . 1 ) of memory block MB 1 and memory transistor cell MC( 0 . 4 ) of memory block MB 2 will be described.

Referring to FIG. 4, in the operation of writing to the right bits, main bit lines MBL 0 , MBL 2 first rise to 5 V. The other main bit lines MBL 1 , MBL 3 are at 0 V. Gate select lines gs 2 , gs 3 then rise to 10 V (“H” level). In response to this, gate transistors gt 2 , gt 6 are turned ON, and the voltage (5 V) on main bit lines MBL 0 , MBL 2 is transmitted to diffusion bit lines BL 2 , BL 6 . In the operation of writing to the right bits, gate select lines gs 1 , gs 4 are at 0 V.

At time t 7 , word line WL 0 is activated and rises to 10 V. In response to this, channel hot electrons are injected to memory transistor cells MC( 0 . 1 ), MC( 0 . 5 ), whereby the data is written to the respective right bits thereof.

After a prescribed time to complete the write operation, that is, at time t 8 , the voltages of main bit lines MBL 0 , MBL 2 are caused to fall.

In response to this, the voltages of diffusion bit lines BL 2 , BL 6 also fall. Once main bit lines MB 0 , MBL 2 and diffusion bit lines BL 2 , BL 6 fall to 0 V at time t 9 , the voltages of gate select lines gs 3 , gs 3 and word line WL 0 are caused to fall. Diffusion bit lines BL 1 , BL 5 are now in the high impedance state.

Adjacent two memory blocks MB are completely isolated from each other by the boundary region therebetween. Therefore, in the above write operation as well, no through current path will be generated between memory blocks MB 1 , MB 2 . As a result, the data can be prevented from being erroneously written to an unintended memory transistor cell MC.

Hereinafter, erase operation will be described.

The erase operation is commonly performed on a block-by-block basis. Therefore, operation of erasing memory block MB 1 will now be described.

Referring to FIG. 5, main bit lines MBL 0 , MBL 1 first rise to 10 V. At time t 10 , gate select lines gs 1 to gs 4 rise to 15 V. In response to this, all gate transistors gt 1 to gt 4 connected, repsectively, to gate select lines gs 1 to gs 4 are turned ON.

As a result, the voltage (10 V) on main bit lines MBL 0 , MBL 1 is supplied to diffusion bit lines BL 0 to BL 3 . In response to supply of such a high voltage to diffusion bit lines BL 0 to BL 3 , electrons trapped in the nitride film of each memory transistor cell MC are removed, whereby the data is erased. After a prescribed time to complete the erase operation, i.e., at time t 11 , main bit lines MBL 0 , MBL 1 fall to 0 V. The voltage levels of diffusion bit lines BL 0 to BL 3 fall in response to the fall of main bit lines MBL 0 , MBL 1 . Once the voltage levels of diffusion bit lines BL 0 to BL 3 fall sufficiently at time t 12 , gate select lines gs 1 to gs 4 are caused to fall. Note that diffusion bit lines BL 0 to BL 3 are now in the high impedance state.

The above erase method enables high-speed batch erase operation.

Second Embodiment

Referring to FIG. 6, a memory array 1100 of the second embodiment is different from memory array 1001 of the first embodiment in that each memory block MB includes sub memory blocks SMB.

The second embodiment provides the structure capable of performing write and read operations one memory transistor cell MC per sub memory block.

Memory block MB 0 includes a sub memory block SMB 1 , a sub memory block SMB 2 , and a memory shield MS. Since sub memory blocks SMB 1 , SMB 2 have the same structure as that of memory blocks MB 1 , MB 2 of the first embodiment, detailed description thereof will not be repeated. Note that sub memory blocks SMB 1 , SMB 2 are herein generally referred to as sub memory blocks SMB.

Memory shield MS is arranged between sub memory blocks SMB 1 , SMB 2 . Memory shield MS includes memory transistor cells MW 0 to MW 31 (hereinafter, generally referred to as memory transistor cells MW). The gates of memory transistor cells MW are connected to word lines WL 0 to WL 31 , respectively. One diffusion region of each memory transistor cell MW is connected to diffusion bit line BL 4 , and the other diffusion region is connected to diffusion bit line BL 3 .

By performing programming operation applying a higher voltage than that of the normal write operation to memory shield MS, a larger amount of electrons are injected thereto. This allows memory transistor cells MW to have a higher threshold voltage than that of normal memory transistor cells MC. For example, in the operation of simultaneously reading or writing the data from or to sub memory blocks SMB 1 , SMB 2 , this higher threshold voltage will not allow memory transistor cells MW to be turned ON even when normal memory transistor cells MC are turned ON.

Arranging memory shield MS be performed the programming operation having a write state higher than the normal write operation between sub memory blocks SMB 1 , SMB 2 prevents generation of a through current path between sub memory blocks SMB 1 , SMB 2 . As a result, erroneous read and write operations can be avoided.

Note that memory shield MS can be formed by injecting a larger amount of electrons than that of the normal write state by performing the programming operation in the shipment.

Hereinafter, erase operation will be described.

Referring to FIG. 7, an erase command is first applied (step S 10 ). An erase pulse having a prescribed pulse width is applied to the memory transistor cells, whereby data is erased (step S 11 ). By using an erase verify function, it is determined whether the data has been erased or not (step S 12 ). If it is determined that the data has not been erased from the memory transistor cells (NO in step S 12 ), it returns to step S 11 . Application of the erase pulse and verification of the erase operation (steps S 11 , S 12 ) are repeatedly performed until the threshold voltage of the memory transistor cells becomes equal to or lower than the reference voltage for determining that the data has been erased. The above process is repeated until the threshold voltage of every memory transistor cell becomes equal to or lower than that reference voltage, that is, until it is determined that the data has been erased from every memory transistor cell.

Thereafter, reprogramming operation is performed to memory shield MS, that is, data is rewritten to each memory transistor cell MW of memory shield MS by using channel hot electrons (step S 13 ). Memory shield MS usually prevents generation of a through current path because of its high threshold voltage. However, since the erase operation is performed on a memory-block by memory-block basis, the injected electrons are removed. As a result, the threshold voltage of memory shield MS may possibly be reduced. The above rewrite sequence which performs the reprogramming operation is therefore provided in order to inject the removed electrons to memory shield MS. This ensures a constant, high threshold voltage of memory shield MS, whereby generation of a through current path can be reliably prevented. Note that, if generation of a through current path cannot be avoided by the same write operation as that of the normal memory transistor cells, the rewrite operation of memory shield MS is performed by applying a voltage to memory shield MS longer than in the write operation of the normal memory transistor cells. Alternatively, a voltage higher than that in the write operation of the normal memory transistor cells may be applied to memory shield MS. As a result, the threshold voltage of memory shield MS becomes high enough to reliably prevent generation of a through current path.

After the reprogramming operation, the erase operation is completed (step S 14 ).

In the erase sequence, the time required for steps S 11 , S 12 generally accounts for a large percentage of the total erase time. Therefore, adding the rewrite step S 13 will not significantly affect the total erase time.

According to the above structure, memory block MB is divided into a plurality of sub memory blocks SMB by memory shield MS. This enables normal write and read operations to be performed on a sub-memory-block by sub-memory-block basis.

As a result, the total number of successively repeated memory-block patterns can be reduced, facilitating configuration of the array.

Third Embodiment

Referring to FIG. 8, an NROM(R) memory array 1200 of the third embodiment is different from NROM(R) memory array 1000 of the first embodiment in that NROM(R) memory array 1200 further includes a memory shield MS between adjacent two memory blocks MB.

This structure enables generation of a through current path to be avoided by memory shield MS when write and read operations are performed simultaneously in adjacent memory blocks MB 1 , MB 2 as described in the first embodiment.

Accordingly, the same effects as those of the first embodiment can be obtained.

Note that the above structure requires the erase sequence described in the second embodiment. More specifically, the reprogramming operation is performed to memory shields MS located on both sides of erased memory block MB.

Physically, every column memory unit MU in the array structure of the third embodiment shares a diffusion bit line with an adjacent column memory unit MU.

Accordingly, it is possible to form memory shields MS by the programming operation each corresponding to a column memory unit MU having a higher threshold value than the normal column memory units MU, and to divide a single memory array into memory blocks MB by memory shields MS, both before shipment.

This facilitates configuration of the memory array in a more flexible manner.

Fourth Embodiment

In the above embodiments, the memory array is divided in order to avoid generation of a through current path. The structure of the fourth embodiment allows the arrangement of memory shields MS described in the second and third embodiments to be changed after shipment.

Memory shields MS described in the second and third embodiments are formed by performing programming operation before shipment. It is normally impossible to access memory shields MS after shipment because addresses are assigned to the memory transistor cells other than those of memory shields MS.

Referring to FIG. 9, a non-volatile semiconductor memory device 2000 is different from non-volatile semiconductor memory device 1000 of the first embodiment in that non-volatile semiconductor memory device 2000 further includes an address terminal 1021 for changing the memory shields, and in that address input buffer 1008 further includes an internal address converter circuit 1020 . Since the circuit structure is otherwise the same as that of the first embodiment, description thereof will not be repeated.

Internal address converter circuit 1020 is a circuit for generating an internal address signal for selecting a column of the array, in response to a column address in the entire memory array including a column address of memory shield MS.

Hereinafter, how memory shield MS is accessed after shipment will be described.

Address input buffer 1008 normally generates an internal address signal in response to an address signal received through data/address terminal group 1006 . More specifically, address input buffer 1008 generates an internal address signal designating a column to be selected from L column memory units of each memory block MB (sub memory block SMB) (where L is a natural number). Address input buffer 1008 then outputs the generated internal address signal to Y decoder 1005 .

In order to change the position of memory shield MS, a memory shield change command is applied from control terminal group 1009 . In response to an address signal for selecting a column in the entire memory array including a column address of memory shield MS from address terminal 1021 , internal address converter circuit 1020 of address input buffer 1008 generates an internal address signal designating a column to be selected. More specifically, internal address converter circuit 1020 generates an internal address signal designating a column to be selected from M column memory units successively arranged in columns in the entire memory array (where M>L; M is a natural number). Address input buffer 1008 then outputs the generated internal address signal to Y decoder 1005 .

As a result, an address is assigned even to memory shield MS by the internal address signal. This enables Y decoder 1005 to access memory shield MS in order to change the position of memory shield MS.

Accessed memory shield MS is thus changed to a normal column memory unit when data is erased therefrom. Moreover, a normal column memory unit can be changed to a memory shield MS having a high threshold voltage by injecting a larger amount of electrons by performing change programming operation. That is, as described in the second embodiment, this injection of electrons is performed by applying a higher voltage than that in the normal write operation or applying a voltage longer than that in the normal write operation.

The structure of the present embodiment enables arrangement of memory shields MS to be changed even after shipment, and therefore allows the user to change the design freely after shipment.

Although the present invention has been described and illustrated in detail, it is clearly understood that the same is by way of illustration and example only and is not to be taken by way of limitation, the sprit and scope of the present invention being limited only by the terms of the appended claims.