This application claims priority/priorities from Japanese Patent Application No. 2012-111974 filed on May 15, 2012, the entire contents of which are incorporated herein by reference.
Embodiments described herein relate generally to a semiconductor integrated circuit.
In recent years, reconfigurable integrated circuit devices such as field programmable gate arrays (FPGAs) have attracted attention. An FPGA includes logic blocks for implementing basic logic information and switches for changing the interconnection between the logic blocks, thereby allowing the user to realize arbitrary logic functions. Logic information implemented by the logic blocks and connection information of the switches are stored in a configuration memory.
As the configuration memory, a static random access memory (SRAM) type memory is widely used. In SRAM-type memories, leakage current increases as the miniaturization of semiconductor devices proceeds. Thus, in FPGA using such SRAM-type memories, leakage current also increases. In order to reduce power consumption, there is considered, for example, a method of shutting off a power supply when the FPGA is in an unused state. However, since the SRAM-type memory is a volatile memory, when the power supply is shut off, data stored in the SRAM-type memory is lost. Thus, the method can not be applied to the FPGA using the SRAM-type memory.
There is also considered a method of mounting a nonvolatile memory on the chip and transferring data to a retaining circuit for retaining data when FPGA recovers from the power-supply shut-off state. However, in this method, nonvolatile memory areas are collectively provided as a block, and the area of the chip increases. In addition, since SRAM or a flip-flop is used as the retaining circuit, the area of the chip further increases.
One embodiment provides a semiconductor integrated circuit, including: a substrate; a plurality of nonvolatile memory portions formed in the substrate, each nonvolatile memory portion including a first nonvolatile memory having a control gate, and a second nonvolatile memory having a control gate connected to the control gate of the second nonvolatile memory; and a plurality of logic transistor portions formed in the substrate, each logic transistor portion including at least one of logic transistor, wherein the logic transistors include: a first transistor which is directly connected to drains of the first and second nonvolatile memories at its gate; and a second transistor which is not directly connected to the drains of the first and second nonvolatile memories, and wherein, among the logic transistors, a bottom surface of the gate of each of the transistors sandwiching the first and second nonvolatile memories is lower in height from a top surface of the substrate than a bottom surface of the control gate of each of the first and second nonvolatile memories.
Hereinafter, an embodiment is described with reference to the drawings. FIG. 1 illustrates one cell of a programmable logic switch (logic switch) according to the present embodiment. A cell 1a of the logic switch according to the present embodiment includes two nonvolatile memories M1 and M2 and one pass transistor PT1. Each of the nonvolatile memories M1 and M2 includes a signal electrode and a control electrode and changes a state thereof according to signals respectively input to the signal electrode and the control electrode. Terminals of the nonvolatile memories M1 and M2 are connected to a node Q. The control electrodes the nonvolatile memories M1 and M2 are connected to a word line WL1. In addition, the nonvolatile memory M1 is connected to a bit line BL1. The nonvolatile memory M2 is connected to a bit line BL2. The gate of the pass transistor PT1 is connected to the node Q.
In the cell 1a, data can be written to one of the nonvolatile memories M1 and M2 by adjusting the magnitude of a voltage applied to each of the word line WL1 and the bit lines BL1 and BL2 and the voltage application timing thereto. In addition, data can collectively be erased from the nonvolatile memories M1 and M2. The nonvolatile memories M1 and M2 are, e.g., flash memory transistors or three-terminal nonvolatile microelectromechanical system (MEMS) switches. Hereinafter, a flash memory transistor having a charge storage film is described as an example for each of the nonvolatile memories M1 and M2.
The memory transistors M11 and M21 may be floating-gate (FG) type transistors using electrically conductive floating gates (e.g., n-doped polysilicon or p-doped polysilicon) as charge storage films. Alternatively, the memory transistors M11 and M21 may be metal-oxide-nitride-oxide-semiconductor (MONOS) type transistors using insulating silicon nitride or oxide films as charge storage films. A structure in which an insulating silicon nitride or oxide film is stacked on an electrically conductive polysilicon may be used. In the case of using the MONOS type transistors as the memory transistors, a manufacturing-process for the logic transistor such as the pass transistor PT1 is compatible with a manufacturing-process for the memory transistors. Thus, the memory transistors can be arranged close to the logic transistor, as compared with the case of using the FG-type transistors as the memory transistors. Accordingly, the area of a chip can be reduced. The following description is made by assuming that the memory transistors M11 and M21 according to the present embodiment are MONOS-type transistors using silicon nitride films as the charge storage films.
Thus, if all of contacts/vias and the gates of transistors are laid out in terms of an equal distance length F when information of 1 bit is represented using the two memory transistors M11 and M21, the minimum area of each object including a margin with each adjacent cell therefrom is 10F×7F, as illustrated in
As illustrated in
Since the gates of the memory transistors M11 and M21 are connected to the word line WL1, the gates of the memory transistors M11 and M21 are at the same voltage. Even in a case where the gate of the memory transistor M11 and the gate of the memory transistor M21 are locally connected to different wires WL1 and WL2, respectively, as illustrated in
The following description is made by assuming that the memory transistors M11 and M21 are N-type transistors formed on a P-type well. However, the memory transistors M11 and M21 may be P-type transistors formed on an N-type well. Although the memory transistors M11 and M21 are formed on the same active area (AA) surrounded by the STI 17, as illustrated in
Hereinafter, a writing method for writing data to the memory transistors M11 and M21 of the cell 1b is described with reference to
The write method according to the present embodiment utilizes channel hot electrons (CHEs). CHEs are electrons each having extremely high energy, and such CHEs are generated when the drain voltage of a transistor becomes larger than a certain amount to thereby cause channel pinch-off at the drain end. The CHEs generated due to the voltage difference between the source and the drain of the transistor are drawn into the charge storage film by a gate voltage. Thus, writing data to the memory transistors is accomplished. The writing method using CHEs has an advantage over a writing method using Fowler-Nordheim (FN) current in that a voltage applied to each terminal is small.
In a case where data is written only to the memory transistor M21 when each of the memory transistors M11 and M21 is in an erased state, a control circuit applies a first write voltage and a second write voltage to the word line WL1 and the word line WL2, respectively, as illustrated in
The first write voltage is set to be equal to or higher than the threshold voltage Vth for the memory transistors M11 and M21 in the erased state, and to be lower than a voltage at which data is written to the memory transistor using FN current. More specifically, a voltage of about 20 V is needed to write data to the memory transistor using FN current. Thus, the first write voltage is set to be lower than 20 V.
The second write voltage is set to be enough to overcome a barrier height between silicon and the tunnel film 51. This is because it is necessary that CHEs generated at the drain end of the memory transistor M21 leap an energy barrier of the silicon oxide film (tunnel film) 51 and go into the silicon nitride film (charge storage film) 12. In addition, the second write voltage is set sufficient not to break the p-n junction formed between a (P-type) well of the memory transistor M21 and an (N-type) source-drain diffusion layer.
In the present embodiment, the first write voltage is set at, e.g., 10 V. The second write voltage is set at, e.g., 4 V. However, the first write voltage and the second write voltage are not limited thereto. The first write voltage and the second write voltage may be the same voltage. The second write voltage may be larger than the first write voltage.
In the writing method, data can be selectively written to the memories, while the breakdown of the gate of the pass transistor PT1 due to the write voltage can be avoided. First, a principle for selectively writing data to the memory is described hereinafter.
When the first write voltage (e.g., 10 V) is applied to the word line WL1, both of the memory transistors M11 and M21 are put into an on-state. The first write voltage is lower than a voltage needed to write data to the memory using FN current. Thus, only the application of the first write voltage thereto is not enough to perform the writing of data to one of the memory transistors M11 and M21.
Then, the second write voltage (e.g., 4 V) is applied to the bit line BL2. Consequently, a voltage difference (drive voltage) between the source and the gate is differentiated between the memory transistors M11 and M21. The drive voltage of the memory transistor M11 is 10 V. The drive voltage of the memory transistor M21 is 6 V. Generally, as the drive voltage of a transistor increases, the channel resistance of the transistor decreases. That is, the memory transistor M11 is lower in resistance than the memory transistor M21 connected to the bit line BL2.
Hereinafter, the channel resistance of the memory transistor M11 is designated with “R1”, while the channel resistance of the memory transistor M21 is designated with “R2”. The voltages respectively applied to the bit lines BL1 and BL2 are designated with “VBL1” and “VBL2”. The voltage at the node Q is designated with “VQ”. The voltage VQ is expressed by the following expression.
VQ=(R1/(R1+R2))*(VBL2−VBL1)
Because R1<R2, the voltage VQ is closer to VBL1 than VBL2. The source-drain voltage of the memory transistor M21 is larger than the source-drain voltage of the memory transistor M11.
As described above, CHEs are generated when the channel of a transistor is pinched off at the drain end thereof. If the source-drain voltage of the memory transistor is increased, channel pinch-off occurs when the source-drain voltage thereof reaches a certain voltage (“VDsat”). If channel pinch-off occurs once, the drain current thereof is saturated.
Next, a principle for avoiding the breakdown of the gate of the pass transistor PT1 due to the write voltage is described hereinafter. In order to implement a high-speed logic switch, preferably, the thickness of a gate insulating film of the pass transistor PT1 is several nanometers (nm). The breakdown withstanding voltage of the gate insulating film in this case is considered to be about 2 V. Accordingly, if the voltage at the node Q is equal to or higher than 2 V, the breakdown of the gate of the pass transistor PT1 may occur.
In the writing method of the present embodiment, the memory transistor M11 is larger than the memory transistor M21 in the drive voltage when writing data. Accordingly, the voltage VQ at the node Q is substantially 0 V. Consequently, the cell 1b is stable in a state in which substantially no voltage is applied to the gate of the pass transistor PT1.
Thus, according to the writing method of the present embodiment, the voltage VQ can be suppressed to 0 V or a value close to 0 V, a voltage that may cause breakdown of the pass transistor PT1 (about 2 V) is not applied to the pass transistor PT1. Accordingly, in the cell 1b, the thickness of the gate insulating film of the pass transistor PT1 can be made sufficiently thin. Thus, a high-speed operation of a logic switch can be accomplished.
As illustrated in
On the other hand, if the first write voltage is given to the word line WL1 before the second write voltage is given to the bit line BL2, a difference in drive voltage is necessarily caused between the memory transistors M11 and M21 when the second write voltage is given to the bit line BL2. The difference in the drive voltage is sufficiently larger than an amount of variation of the threshold voltage of the memory transistor. Thus, when the second write voltage is given to the bit line BL2, the voltage VQ does not become unexpectedly close to the voltage VBL2. A voltage more than expected is not applied to the pass transistor PT1.
In the case of arranging cells according to the present embodiment like an array, plural cells are connected to the same word line, as illustrated in
As illustrated in
According to the present embodiment, there is no need to newly add another element to a cell for the purposes of selectively writing data to the memory transistor and of preventing the breakdown of the gate insulating film of the pass transistor. In addition, the two memory transistors included in each cell are connected to one of the word lines. Thus, according to the present embodiment, a programmable logic switch is implemented, which has a small chip area and which can perform the selective writing of data to the memory transistor so as to prevent a high voltage from being applied to the pass transistor due to the writing of data to the memory transistor.
As illustrated in
Incidentally, in the above description, the writing method for writing data to the memory transistor M21 has been described. Writing data to the memory transistor M11 can be accomplished by a similar method. In the case of writing data to the memory transistor M11, the second write voltage is applied to the bit line BL1. In addition, the bit line BL2 is set to the ground voltage.
Although the ground voltage is given to the bit line BL1 and the substrate electrode in the above description, a third write voltage, whose sign is negative, may be given thereto. Increase of electron injection efficiency and reduction of the first write voltage can be expected by setting the voltage of the well of the memory transistor to be negative. However, the third write voltage should be set such that the gate insulating film of the pass transistor is prevented from being broken down by the third write voltage applied to the substrate electrode. The voltage at the node Q, which is obtained when writing data, is substantially equal to the voltage VBL1. Thus, it is necessary that the difference between the voltage given to the bit line BL1 and the third write voltage is equal to or less than the withstanding voltage of the gate insulating film of the pass transistor. For example, if the third write voltage is also given to the bit line BL1, a load given to the gate insulating film of the pass transistor can be minimized. The voltage given to the bit line BL1 and the voltage given to the substrate electrode may be either equal to or different from each other.
Hereinafter, a method for erasing data from the memory transistor according to the present embodiment is described with reference to
If the negative erasure voltage is applied to the word line WL1, it is expected that the voltage VQ at the node Q is increased in a negative direction by the capacitive coupling between the gate and the drain of the memory transistor. However, if the voltage VQ is lower than the substrate voltage, current flows through the p-n junction formed between the (P-doped) well of the memory transistor and the (N-doped) source-drain diffusion layer. Thus, the voltage VQ is immediately settled to the same voltage as the substrate voltage. Accordingly, the application of a high voltage to the pass transistor PT1 by the capacitive coupling of the erasure voltage does not occur.
Instead of the above-mentioned erasure method, a positive erasure voltage (e.g., about 20 V) may be applied to the substrate electrode, and a voltage of 0 V may be given to the word line WL1. However, according to the present embodiment, the well is common to the memory transistors M11 and M21 and the pass transistor PT1. Thus, when the erasure voltage is given to the memory transistors M11 and M21, the erasure voltage is also applied to the pass transistor PT1, and, the erasure voltage is output from the pass transistor PT1 through the p-n junction formed between the source/drain and the well thereof. The source/drain of the pass transistor PT1 may be connected to another logic circuit (e.g., the input and output terminals of an inverter), and a transistor of the another logic circuit may have the thin gate insulating film for realizing high-speed operation. When the erasure voltage serving as a high voltage output from the pass transistor PT1 is applied to the another logic circuit, the breakdown of the gate insulating film may be caused in the transistor of the another logic circuit.
On the other hand, if the erasure method according to the present embodiment is used, the voltages of the source and the drain of the pass transistor are not caused to be high. Thus, the logic circuit connected to the pass transistor is allowed to use a high-speed transistor having a thin gate oxide film.
As an operation method of the logic switch according to the present embodiment, the control circuit applies, to the word line WL1, a voltage intermediate between the threshold voltages respectively corresponding to the erased state and the written state. Also, the control circuit applies a first operating voltage and a second operating voltage which is smaller than the first operating voltage to the bit lines BL1 and BL2, respectively. Consequently, the memory transistor in the erased state is put into an on-state, while the memory transistor in the written state is put into an off-state. If the pass transistor is an N-type transistor, the pass transistor becomes an on-state by receiving the first operating voltage applied through the memory transistor in the erased state at its gate, and becomes an off-state by receiving the second operating voltage at its gate. In the operation method, when the voltage applied to the gate of the pass transistor is set to be higher than the voltage of a signal input to the source/drain of the pass transistor, a signal passing through the pass transistor fully swings. Thus, power consumption is reduced, and a delay is decreased. More specifically, the value of the voltage applied to the gate of the pass transistor in an on-state is set to be larger than a value obtained by adding the value of the voltage of the signal to the value of the threshold voltage of the pass transistor.
The value of the voltage applied to the gate of the pass transistor in an on-state is set in the following manner. That is, first, the first operating voltage is given thereto via the memory transistor which is in an erased state. In order to give the first operating voltage to the gate of the pass transistor in a case where the memory transistor is an N-type transistor, it is necessary to set the value of the voltage given to the gate of the memory transistor as being larger than a value obtained by adding the first operating voltage and the threshold value for the pass transistor.
Here, the higher-side voltage of a signal passing through the pass transistor is designated with Vdh, the lower-side voltage thereof is designated with Vdl. The threshold voltage of the pass transistor is designated with Vthpt, and the threshold voltage of the N-type memory transistor in an erased state is designated with Vthm. A voltage which is applied to the word line WL1 and which is a voltage intermediate between the threshold voltages respectively corresponding to the erased state and the written state is designated with Vw1. And, the first operating voltage is designated with V1, and the second operating voltage is designated with V2. In order to give the first operating voltage and the second operating voltage via the memory transistors, it is necessary to satisfy the following relations:
Vw1>Vthm+V1; and
Vw1>Vthm+V2.
Assuming that the first operating voltage is larger than the second operating voltage, it is sufficient to satisfy only the above relation:
Vw1>Vthm+V1.
In addition, considering conditions for allowing, when the pass transistor is an on-state, the signal passing the pass transistor to fully swing, and for inhibiting, when the pass transistor is an off-state, a signal from passing through the pass transistor, it is necessary to satisfy the following conditions:
V1>Vdh+Vthpt; and
V2<Vdl+Vthpt.
In addition, the following relation is derived from the above expressions:
Vw1>Vthm+Vdh+Vthpt.
If the voltages satisfy the above relations, the power consumption is not reduced, and the reduction of the delay is suppressed. In a case where the pass transistor is a P-type transistor, the above expressions can also be represented by similar expressions by setting the threshold voltage Vthpt at a negative value.
If the above-mentioned layout, the number of wires provided on the chip or a total wiring length can be reduced, so that the chip area can be decreased, because the distance between each memory transistor and the logic transistor can be short, as compared with the layout in which the memory transistors are gathered at one place. Consequently, the chip area can be reduced. If the power supply to the logic switch is resumed after it is once interrupted, the on/off of the pass transistor PT1 can quickly be controlled according to data stored in the memory transistors M11 and M12. Thus, power-shutdown and power-recovery can immediately be achieved. In the conventional technique, since electric-power is supplied through a long wire, power-recovery takes much time, as compared with the present embodiment.
The semiconductor integrated circuit having the logic switch according to the present embodiment may employ layouts respectively illustrated in
In the layouts illustrated in
In the case of arranging the transistors, as illustrated in
The semiconductor integrated circuit having the logic switch according to the present embodiment may employ a layout as illustrated in
In each of the layouts illustrated in
Each of
If an affection from the adjacent memory transistors is small, the memory transistors can be arranged in a high density by reducing the distance between the memory transistors provided in each cell, as illustrated in
In the case of the layout of the memory transistors illustrated in
Hereinafter, the channel lengths and the channel widths of the memory transistors and the logic transistor are described with reference to
Such different channel lengths have lengths that differ from each other by amounts larger than a manufacturing variation. According to, e.g., International Technology Roadmap for Semiconductors (ITRS), channel lengths one of which differs from the other in length by 10% or more are referred to as “different channel lengths”.
Concerning the channel widths of the memory transistors and the logic transistor, each memory transistor may be larger than the logic transistor in channel width, as illustrated in
Thus, according to the present embodiment, a cell of the logic switch includes two nonvolatile memory transistors that the gate is connected to a single word line, and a logic transistor the gate of which is connected to a node at which the source end or the drain end of one of the two memory transistors is connected to the drain end or the source end of the other memory transistor. In addition, regions at each of which an associated one of the memory transistors is provided are separately arranged in a chip. The employment of nonvolatile memory transistors enables the shutdown of electric-power. Thus, power consumption can be reduced. In addition, there is no need to provide two memory transistors such that the source end or the drain end of one of the two memory transistors is connected to a node to which the drain end or the source end of the other memory transistor is connected, in order to perform the selective writing of data to the memory transistors. In addition, in order to arrange each memory transistor and the logic transistor closely to each other, a total wiring length can be reduced. Consequently, the chip area can be decreased.
In
A switch 202 includes a complementary metal-oxide semiconductor (CMOS) inverter, an N-type pass transistor switch connected to a ground-side terminal of the CMOS inverter, and an N-type pass transistor switch connected an output terminal of the CMOS inverter. When the CMOS inverter is used, although an output direction is limited to one direction, the drivability of the wires is enhanced. Thus a wiring delay is reduced. The on/off of each of the switches respectively connected to the ground-side terminal and the output terminal of the CMOS inverter is controlled according to the value stored in the memory transistor. If the pass transistor switch connected to the output terminal of the CMOS inverter is turned off, the power supply to the CMOS inverter is shut down. Thus, the power consumption of the CMOS inverter can be reduced.
A switch 203 includes, in addition to components (a CMOS inverter and an N-type pass transistor) almost similar to that of the switch 202, a NAND gate connected to an input-side terminal of the CMOS inverter. An output of the corresponding memory transistor is connected to one of the input terminals of the NAND gate. Although a transistor (N-type pass transistor) is not connected to the ground-side terminal of the CMOS inverter in the switch 203 unlike the switch 202, a power-off transistor may be provided similarly to the switch 202. While a logical value of a signal is inverted in the switch 202 because it has only one inverter, in the switch 203, the inversion of the logical value of the signal is not caused. Accordingly, the design of the device is facilitated. If the pass transistor switch is turned off, the power supply to the ground-side terminal of the NAND gate is interrupted. In addition, a P-type transistor at the power-supply voltage side of the NAND gate is turned on. Thus, a signal of a High level is input to the input terminal of the CMOS inverter. Generally, when the voltage at an input terminal of a CMOS inverter is an intermediate value that is neither High level nor Low level, a leakage current of the CMOS inverter will increase. However, according to the switch 203, the increase of the leakage current can be suppressed by shutting down the power supply. At switching in an FPGA operation, no current flows in the P-type transistor of the NAND gate, which is connected to the memory transistor. Thus, the channel length of the P-type transistor may be increased. The channel width thereof may be reduced. Alternatively, plural P-type transistors may be series-connected to the memory transistor. Consequently, when the pass transistor is turned off, the leakage current of the NAND gate itself can be reduced when the pass transistor is turned off.
A switch 204 has a structure almost similar to that of the switch 202. The pass transistor thereof is connected to an output terminal of the CMOS inverter. However, an output terminal of the memory transistor is directly connected to an input terminal of the CMOS inverter. Another signal other than the output signal from the memory transistor is connected to the gate of the pass transistor. Consequently, a logical operation can be performed using a signal other than the output signal from the memory transistor, thereby forming, e.g., a look-up table (LUT).
In the case of any of the above switches, an output of the memory transistor is connected only to the gate electrode of the logic transistor. In the foregoing description, the case of using the N-type transistor as the pass transistor is exemplified. In the case of using a P-type transistor as the pass transistor, the switch will be connected to the memory transistor by, e.g., appropriately performing a logical inversion, especially, in the switches 202 and 203, so that the power supply to the CMOS inverter and the NAND gate is interrupted when the pass transistor is turned off.
In the case of using a MONOS type transistor as the memory transistor according to the present embodiment, preferably, in the charge storage film, a deviation of electric-charge injected thereinto is small, as will be described below.
The electric charge trap level of silicon nitride (SiN) used in the charge storage film 52 of the present embodiment differs according to the content rates of Si and N. Assuming that the proportion of a molar ratio of Si in SiN to a molar ratio of N is referred to as N/Si ratio, a Si atom has four unpaired electrons, and an N atom has three unpaired electrons. Therefore, a stoichiometric composition ratio N/Si of N and Si is 1.33 (N/Si=1.33). A SiN film whose N/Si ratio is 1.33 is referred to as a stoichiometric SiN film, and a SiN film whose N/Si ratio is smaller than that of the stoichiometric SiN film is referred to as a Si-rich SiN film. Here, the Si-rich SiN film is shallower in electron trap level than the stoichiometric SiN film. Thus, electrons in the film are relatively easy to move.
If a stoichiometric SiN film is used as the charge storage film 52, electrons injected into the charge storage film locally exist and are trapped by writing data. When writing data, CHEs are generated at the drain end of the memory transistor. Thus, electrons trapped in the charge storage film by writing data using CHEs are concentrated at the drain side. Accordingly, asymmetry may occur in the voltage distribution in the channel of the memory transistor.
In an n-channel transistor, one of source-drain diffusion regions, which is higher in voltage than the other source-drain diffusion region, is defined as a “drain”. The other source-drain diffusion region which is lower in voltage than the “drain” is defined as a “source”. Thus, the channel resistance of the transistor is largely dominated by a source-side voltage barrier. For example, it is assumed that electrons of the same amount are injected into each of the charge storage films of the two memory transistors, that a larger amount of electrons are stored in a drain-side part of the charge storage film of one of the memory transistors, and that a larger amount of electrons are stored in a source-side part of the charge storage film of the other memory transistor. The memory transistor adapted to store a larger amount of electrons in the source-side part of the charge storage film can strongly modulate the voltage of the source, so that change of the threshold voltage Vth is increased (see, e.g., IEEE ELECTRON DEVICE LETTERS, Vol. 21, pp. 543 to 545, 2000).
Thus, it is considered that when electrons locally exist in the charge storage film in the vicinity of the drain due to the writing using CHEs and are trapped, the threshold voltage Vth is not sufficiently changed in some source-drain voltage application direction.
If the threshold voltage Vth is not sufficiently changed, the logic switch may not be appropriately operated. The case of applying a voltage to each wire of the cell 1b to operate the logic switch, as illustrated in
Thus, it is preferable that when data is written to the memory transistor, electrons are caused to distribute in the source-side part of the charge storage film. However, in the case of using a stoichiometric SiN film as the charge storage film, it is necessary to increase a write time in order to implement, on the source-side part of the charge storage film, the distribution of the electrons injected when the memory transistor is put into a written state.
On the other hand, in the case of using a Si-rich SiN film as the charge storage film 52, the trap level of the Si-rich SiN film is shallow, as compared with the trap level of the stoichiometric SiN film. Thus, electrons are easy to move in the Si-rich SiN film. Accordingly, electrons injected into the drain end by writing data move in the charge storage film 52 and are diffused to the source end. Thus, if the logic switch is operated by applying a voltage to each wire of the cell 1b, as illustrated in
As the N/Si ratio decreases, the electrons injected into the charge storage film become more easily to move. However, when the N/Si ratio reaches 0.67, since this ratio corresponds to a composition in which averagely, two of four bonds of Si are converted into dangling bonds, or in which adjacent Si atoms form a covalent bond, there will be a large number of covalent bonds. Thus, the insulation property of the SiN film is deteriorated, so that gate leakage current extremely increases. Accordingly, it is desirable that the N/Si ratio is larger than 0.67 and smaller than 1.33. The composition of the charge storage film can be known by being analyzed according to an electron energy-loss spectroscopy (EELS).
In the case of using a Si-rich SiN film as the charge storage film, because the trap level is shallow, it is frequent that the trapped electrons go out of the charge storage film to the substrate via the tunnel film 51 or to the gate electrode 54 via the block film 53 due to energy obtained from heat. This means that a data retention time of the memory transistor is shorten. A proportion of the electrons caused to go out to the substrate is larger than that of the electrons caused to go out to the gate electrode 54. Therefore, in order to improve the data retention time of the memory transistor, it is necessary to prevent electrons trapped in the charge storage film 52 from going out to the substrate side.
Thus, the N/Si ratio of the charge storage film 52 is changed in this film in a stacking direction. More specifically, a portion of the charge storage film 52, which portion is located in the vicinity of the tunnel film 51, is increased in the N/Si ratio in order not to degrade the retention characteristic thereof. On the other hand, another portion of the charge storage film 52, which portion is located in the vicinity of the block film 53, is decreased in the N/Si ratio in order to facilitate the movement of electric-charge in this film. Consequently, an amount of change of the threshold voltage Vth due to the writing of data can be increased. In addition, the data retention time of the memory transistor can be increased. In the case of the memory transistor used as a file memory, the increasing of the N/Si ratio of a portion of the charge storage film, which portion is in the vicinity of the tunnel film 51, is not permitted, because an erasure time is long. However, the memory transistors used in the programmable switch according to the present embodiment is predominantly low in rewriting-frequency, as compared with the memory transistor used as a file memory. Therefore, the increase of the erasure time is not problematic. The increasing of the N/Si ratio of a portion of the charge storage film 52, which portion is in the vicinity of the tunnel film 51, has a great advantage.
Although the case of using a silicon nitride film as the charge storage film 52 has been described, even in the case of using a silicon oxynitride film as the charge storage film 52, electrons can be prevented using a Si-rich silicon oxynitride film from locally existing.
In the case of using MONOS type transistors as the memory transistors according to the present embodiment, it is preferable that the block film 53 of each memory transistor is made of a material or film with low charge permeability. The charge storage film 52 in a typical flash memory exchanges charges to and from the substrate. Therefore, it is not preferable that the charge storage film 52 exchanges electric-charges to and from other portions (such as the gate electrode 54). In the case of applying the memory transistor to the logic switch like the present embodiment, a problem may be caused by injecting electric-charges from the gate electrode 54 to the charge storage film 52 or by discharging electric-charges (back-tunneling) from the charge storage film 52.
For example, in the case of using a SiN film as the charge storage film 52 and employing silicon oxide (e.g., SiO2) or a Si-based material such as SiN and the like as the material of the block film 53, when an erasure operation is performed, a small number of electrons moved from the gate electrode 54 to the charge storage film 52 due to back-tunneling. Thus, reduction of the threshold voltage Vth is inhibited. Accordingly, the threshold voltage Vth for the memory transistor in an erased state becomes equal to or higher than 0 V. When the logic transistor is operated, a voltage (i.e., a read voltage) applied to the gate electrode of the memory transistor should be set to be larger than the threshold voltage Vth corresponding to an erased state. That is, in this case, it is necessary to apply a positive read voltage to the gate electrode. Typically, the read voltage is about 4 V.
When the logic switch operates, it is necessary to always apply the read voltage to the memory transistor. Error writing of data to the memory transistor being in an erased state may be caused by thus always applying a positive read voltage to the memory transistor. Accordingly, a malfunction of the logic switch may occur. Generally, if a voluntary memory is used as a file memory, a read voltage is not applied to the nonvolatile memory in a retention state. However, if the nonvolatile memory is applied to the logic switch, similarly to the present embodiment, long-term reliability in a state of always applying a read voltage to the nonvolatile memory should be ensured.
Thus, back-tunneling is suppressed using the block film 53 with low charge permeability in the memory transistor. Consequently, the long-term reliability of the logic switch is ensured.
A first example of the block film 53 with low charge permeability is a block film made of a highly insulating substance whose relative permittivity with respect to the permittivity of vacuum is higher than the relative permittivity of SiN (i.e., 7.0). The insulating substance is, e.g., aluminum oxide or hafnium oxide. As the permittivity of the block film 53 increases, the electrostatic coupling between the gate electrode and the substrate becomes stronger. Accordingly, even if the physical thickness of the block film 53 is increased, an electric field applied between the gate electrode and the substrate can be maintained at a high level. If the physical thickness of the block film 53 is increased, the efficiency in exchanging electric-charges between the gate electrode 54 and the charge storage film 52 is reduced. Consequently, back-tunneling can be prevented from occurring at the writing of data to and at the erasing of data from the memory transistor.
Thus, the threshold voltage Vth corresponding to the erased state of the memory transistor can be reduced to a negative value by preventing back-tunneling. Consequently, when the logic switch operates, a read voltage applied to the gate electrode 54 of the memory transistor can be set at 0 V. That is, the state of the memory transistor can be prevented from being changed due to the fact that a nonzero read voltage is always applied thereto when the logic switch operates. In addition, because the read voltage can be set at 0 V, there is no need to provide a power supply exclusively used to supply a read voltage.
The threshold voltage can be obtained, e.g., in the following way. That is, first, a voltage of 50 milli-volts (mV) is applied to between the source and the drain of the transistor. Then, a source-drain current IDS is measured while the gate voltage is changed. Assuming that the channel width of the transistor is designated with W, and that the channel length of the transistor is designated with L, a gate voltage, according to which the following relation holds: IDS*L/W=10 nano-amperes (nA), is defined as a threshold voltage.
In addition to the use of the above block film 53, a metal material whose work function has a value larger than a value (4.05 electron volts (eV)) of the work function of high-density n-doped polysilicon is used as the material of the gate electrode 54 of the memory transistor. For example, tantalum, tungsten and titanium nitride are used as such a metal material. As the value of the work function of the gate electrode 54 increases, the electron barrier height of the block film 53 becomes higher, which is evaluated with respect to the gate electrode 54. Thus, back-tunneling can be prevented from occurring while an erasure operation of the memory transistor is performed.
The material of the insulating film 531a is an insulating material which has a permittivity higher than that of SiN and is, e.g., aluminum oxide or hafnium oxide. The material of the insulating film 532a is SiN. A high-permittivity material is used as the material of the insulating film 531a. Thus, the physical thickness of the insulating film 531a can be increased while the level of the electric field applied between the gate electrode and the substrate is maintained. Consequently, occurrence of back-tunneling is prevented by increasing the thickness of the insulating film 531a. The electric field caused at the gate electrode end during erasure of the memory transistor is reduced in magnitude by providing the insulating film 532a made of SiN on the insulating film 531a. Thus, back-tunneling is more suppressed.
Purposes of inserting the insulating film 532b between the insulating films 531b and 533b are to generate an electric dipole due to a high-permittivity material between the insulating films 531b and 533b, and to increase the barrier height of the insulating film 531b on the interface between insulating films 531b and 532b. Consequently, the back-tunneling from the gate electrode 54 to the charge storage film 52 is suppressed from occurring during an erasure operation (see, e.g., K. Kita, “Intrinsic Origin of Electric Dipoles Formed at High-k/SiO2 Interface”, IEEE International Electron Devices Meeting 2008). In addition, the magnitude of the electric field at the gate electrode end, which is caused during an erasure operation of the memory transistor, is reduced by providing the insulating film 533b on the insulating film 532b. Consequently, the effects of back-tunneling are more suppressed.
In the case of inserting the insulating film 532b between the insulating films 531b and 533b like the present embodiment, the thickness of the insulating film 532b can be reduced. Thus, the thickness of the insulating film 532b is reduced to an extremely small value. Consequently, substantially no change of a conventional processing process for memory transistors that use only Si-based materials is needed. In addition, the deterioration of the memory characteristic due to the diffusion of high-permittivity materials into the charge storage film 52 can be prevented by providing the insulating film 531b made of SiO2 between the charge storage film 52 and the insulating film 532b made of a high-permittivity material.
Hereinafter, a manufacturing method for a programmable logic switch according to the present embodiment is described. The following description is made by assuming that each memory transistor is a MONOS-type transistor using silicon nitride as the material of the charge storage film. Briefly, a programmable logic switch according to the present embodiment is manufactured as follows. That is, first, the formation of STI (shallow trench isolation) and the deposition of the insulating films of the memory transistors are performed. Then, the insulating film of the pass transistor is formed. In addition, the processing of the gate electrodes and gate insulating films of the memory transistors and the pass transistor is performed. Then, wiring is performed.
Either of the manufacture of STI and the deposition of the insulating films of the memory transistors may be performed before the other. That is, STI may be manufacture after the insulating films of the memory transistors are deposited. Alternatively, the insulating films of the memory transistors may be deposited after STI is manufactured.
Preferably, the surface of the STI 17 is located higher than parts of the substrate 9, which respectively correspond to the element regions 110 and 111, and lower than the surface of the nitride film 12. Thus, wet-etching is performed (see
Then, an oxide film 13 serving as the block film is formed (see
Thus, in the case of depositing the insulating films of the memory transistors before the manufacture of the STI, etching is performed on a region in which the STI 17 is manufactured, in a state in which the oxide film 11 and the nitride film 12 are stacked. Consequently, as viewed in a gate cross-section in a direction perpendicular to the channel of each memory transistor, end portions in a channel width direction (a process end processed by the dry-etching) of the tunnel film and the charge storage film constitute surfaces continuous to the boundaries between the STI 17 and each of the element regions 110 and 111. Accordingly, the electric field generated between the charge storage film and the substrate becomes uniform. When data is written to the memory transistor, electrons can uniformly be injected into the charge storage film from the substrate.
Preferably, the surface of STI 17 is located higher than the surface of the substrate 9 at each of the element regions 110 and 111. In addition, preferably, the height of the surface of the STI 17 from the surface of the substrate 9 is set to be equal to or less than a sum of the physical thicknesses of the tunnel film, the charge storage film and the block film.
If the surface of the STI 17 is located lower than the surface the substrate 9 at each of the element regions 110 and 111, a corner of the substrate 9 is exposed. Thus, in a process of forming the gate insulating film of the pass transistor, which will be described below, when the substrate is oxidized, oxidization may become insufficient. This may cause increase of gate leakage current.
If the height of the surface of the STI 17 is larger than a sum of the physical thicknesses of the tunnel film, the charge storage film and the block film, the dimensions of the gate patterns may differ from intended dimensions when lithography of the gate patterns of the memory transistors and the pass transistor is performed in a process which will be described below. If the lithography of the gate pattern of each memory transistor and the lithography of the gate pattern of the pass transistor are performed collectively, the height from the substrate to the top surface of each of the materials of the gate electrodes respectively corresponding to the memory transistors and the pass transistor may differ from one another when the materials of gate electrodes are deposited on the element region 111 in which each memory transistor is formed, and the element region 110 in which the pass transistor is formed, respectively, as illustrated in
If a focus height for providing optimum exposure of the element region 111, in which each memory transistor is formed, is set as a reference height, a focus height for providing optimum exposure of the element region 110, in which the pass transistor is formed, is lower than the reference height by a sum of the physical thicknesses of the tunnel film, the charge storage film and the block film of each memory transistor. In addition, a focus height for providing optimum exposure of the STI region 17 is higher than the reference height by the height of the STI. If the height of the surface of the STI 17 is higher than a sum of the physical thicknesses of the tunnel film, the charge storage film and the block film, the difference between the focus height for providing optimum exposure of the STI region 17 and the focus height for providing optimum exposure of the element region 110 is large. Thus, no height may be determined as a focus height for providing both of optimum exposure of the STI region 17 and optimum exposure of the element region 110.
Preferably, a side surface of a part of the STI 17 which projects from the surface of the substrate 9 form an angle with the surface of the substrate 9 which is equal to or more than 90 degrees. Consequently, in the cases of writing data to or erasing data from memory transistors manufactured by the process which will be described below, a strong electric field can be prevented from locally being applied to the tunnel film provided along the boundary between the STI 17 and each of the parts of the substrate 9.
As illustrated in
It is necessary to form the tunnel film, the charge storage film, the block film of each memory transistor and the gate insulating film of the pass transistor as different films. Thus, the insulating film of the pass transistor may be manufactured after all of the oxide film 11, the nitride film 12 and the oxide film 13 in the element region 110 are eliminated. Alternatively, while all of the oxide film 13 and the nitride film 12 of the element region 110 in which the pass transistor is formed are removed, a part or all of the oxide film 11 may be left as the insulating film of the pass transistor.
The gate electrode material 14 may be, e.g., polysilicon, a metal material, or metal and polysilicon layer stacks. High-melting-point metal such as tantalum, titanium, tungsten and molybdenum, or carbides and nitrides of such metal, or aluminum compounds, e.g., TaC, TaN, TiN, TiCN, TiAlN, W, WN and Mo, can be used as the metal material. In the case of using polysilicon as the gate electrode material, ion injection is performed, if necessary.
In the element regions 110 and 111, films are simultaneously formed using the gate electrode materials. Although films can be formed using the gate electrode materials in each of the element regions, if the number of processes for manufacturing gate electrodes increase, interface states are formed in the memory transistors and the pass transistor. Thus the performance of the transistors is degraded. Thus, films are simultaneously formed using the gate electrode materials in the element regions 110 and 111 to thereby minimize the number of processes for manufacturing the gate electrodes. Consequently, the degradation of the reliability of the memory transistors and the reduction of the speed of the pass transistor are suppressed.
A method for eliminating the oxide film 11, the nitride film 12 and the oxide film 13 of the element region 110 may be a method using only dry-etching, a method using only wet-etching, or a method of etching part of the oxide film 13, the nitride film 12 and a part of the oxide film 11 by dry-etching and next etching the rest of the oxide film 11 by wet-etching.
Dry-etching excels in anisotropy. Each of the oxide film 11, the nitride film 12 and the oxide film 13 can accurately be processed into resists, using dry-etching. In addition, dry-etching is effective in chip miniaturization. If an appropriate etching gas is selected, plural films can be processed at once using the same etching-gas. Time taken to processing is short.
Wet-etching excels in selectivity among the substrate (Si), the oxide film 11, the nitride film 12 and the oxide film 13. Thus, as compared with the dry-etching, damage given to the substrate 9 is extremely low. Because the wet-etching uses no vacuum process, etching can be performed at low cost.
In the case of etching the oxide film 13, the nitride film 12 and a part of the oxide film 11 by dry-etching, and then etching the rest of the oxide film 11 by wet-etching, both of the advantages of the dry-etching and the wet-etching can be utilized. That is, because the oxide film 13 and the nitride film 12, which have thicknesses that occupy a large part of thicknesses of the insulating films of the memory transistors, are etched by dry-etching, each of the oxide film 13 and the nitride film 12 can accurately be processed into the shape of a resist-pattern. In addition, because the rest of the oxide film 11 is etched using wet-etching, the substrate can be prevented from being damaged.
In the case of using both of the dry-etching and the wet-etching, if the material of the oxide film 11 is SiO2, it is preferable that a film made of a material, such as SiN or amorphous silicon, differing from SiO2 is formed on the oxide film 13. The reason is that when the rest of the oxide film 11 is etched by wet-etching after a part of the oxide film 11 is subjected to dry-etching, if the material of the topmost surface of the oxide film 13 is the same as the material of the oxide film 11, i.e., SiO2, the thickness of the oxide film 13 is reduced. If the material of the topmost surface of the oxide film 13 is SiO2, it is necessary to protect the oxide film 13 by performing lithography thereon after the dry-etching. Consequently, the number of processes increases.
In addition, it is considered that wet-etching is performed without peeling off a resist used in performing dry-etching. However, after the dry-etching, the remaining resist is denatured by plasmas. Therefore, unless the remaining resist is subjected to ashing or the like, the remaining resist cannot be peeled off. If wet-etching is thus performed on the remaining resist without peeling off the remaining resist, it is necessary to perform, after the dry-etching, ashing to peel off the resist. Consequently, the part of the substrate, which part is in the element region 110, is oxidized by ashing. Accordingly, the property of the pass transistor is deteriorated. On the other hand, if a film whose material differs from SiO2 is formed on the oxide film 13, the part of the substrate 9, which part is included in the element region 110, is protected by a part of the oxide film 11 even when the remaining resist is peeled off by ashing before the wet-etching upon completion of the dry-etching. Even if wet-etching is then performed, because the film whose material differs from SiO2 is formed on the oxide film 13, the thickness of the oxide film 13 can be prevented from being reduced.
It is more preferable that SiN film is further formed on the oxide film 13. SiN is higher in permittivity than SiO2. Thus the magnitude of an electric field generated at the interface between the block film and the gate electrode can be reduced. Due to this effect (“retarding effect”), electric-charge can be prevented from being exchanged between the charge storage film and the gate electrode through the block film. Consequently, the characteristics of the memory transistor are improved. In addition, if a buffered hydrofluoric acid is used as an etchant for wet-etching, only the oxide film made of SiO2 is selectively removed. Consequently, the thickness of the SiN film is substantially unchanged.
However, if ashing is performed to remove resist upon completion of dry-etching, the surface of the SiN film may be partly oxidized by plasmas. The hydrofluoric acid resistance of the oxidized part of the surface of the SiN film is weak. Thus, the oxidized part of the surface of the SiN film is removed by wet-etching. Consequently, the thickness of the SiN film is reduced. In such a case, preliminarily considering the above reduction of the thickness of the SiN film, it is advisable to form the SiN film having a larger thickness. Thus, the characteristic of the memory transistor can be prevented from being degraded. For example, assuming that a part having a depth of 2 nm from the surface of the SiN film would be oxidized by the ashing, and that a finally desired thickness of the SiN film is 3 nm, it is useful to form a SiN film having a thickness of 5 nm.
As illustrated in
The properties of both of the pass transistor and the memory transistor can be improved by forming the high-k film 20. The EOT of the high-k film is thin. Therefore, as compared with the case of using SiO2 film as the gate insulating film, a current driving force can be increased with substantially equal amount of gate leakage current provided, using the high-k film as the gate insulating film of the pass transistor. In addition, the magnitude of the electric field generated at the interface between the block film and the gate electrode is reduced (what is called a retarding effect) by inserting the high-k film 20 into the block film of the memory transistor. Consequently, electric-charge can be prevented from being exchanged between the charge storage film and the gate electrode through the block film. Thus, the characteristics of the memory transistor are improved.
The advantages of the dry-etching and the wet-etching are as described above. In addition, if the wet-etching is used in the case of leaving a part or all of the oxide film 11 unremoved, etching can accurately be stopped at the boundary between two of the different films, because the logical switch excels in the selectivity for the etching of the oxide film 11, the nitride film 12 and the oxide film 13. Thus, variation of the thickness of each of the remaining films after the etching is performed is extremely small. The variation of the thickness of the gate insulating film of the pass transistor is small in a wafer.
After the etching is performed such that a part or all of the oxide film 13 of the element region 110 is left, the gate electrode material 14 is simultaneously deposited on the element regions 110 and 111. A material similar to the material used in the case of removing all of the oxide film 11, the nitride film 12 and the oxide film 13 of the element region 110 can be used as the gate electrode material.
As illustrated in
In
The oxide film 11, the nitride film 12, the oxide film 13 and the oxide film 18 are processed to thereby manufacture the gate electrode, the tunnel film, the charge storage film and the block film of the memory transistor, and the gate electrode and the gate insulating film of the pass transistor. Hereinafter, the tunnel film, the charge storage film and the block film of the memory transistor are generically as memory films.
First, a mask material 31 is deposited on the gate electrode material 14 (see
If the gate pattern of the memory transistor and the gate pattern of the pass transistor are processed by being lithographed, similarly to the first example, conditions for exposure at the time of lithographing of the memory transistors and the pass transistor can be adjusted independent of one another. As described above, in the case of collectively lithographing the gate pattern of the memory transistor and the gate pattern of the passing transistor, it is necessary for reducing the influence of the difference in height between the element regions 110 and 111 to determine a focus height for exposure. In the case of performing exposure of minute gate patterns, a finished pattern is affected by a slight difference in height from the substrate. Optimum exposure of each element region can be performed by adjusting exposure conditions at the time of lithographing of the memory transistors and the pass transistor. Devices can be processed to designed dimensions.
A second example of a process of processing the gate electrode, the gate insulating film and memory films is described hereinafter with reference to
Similarly to the first example, in the second example, first, the mask material 31 is deposited (see
Then, the gate electrode material 14 on the element regions 110 and 111 are collectively etched using the mask material 31 as a mask (see
Similarly to the first example, even in the second example, the gate pattern of each memory transistor and the gate pattern of the pass transistor are processed in different steps by being lithographed. Thus, exposure conditions for lithographing each memory transistor and the pass transistor can be adjusted independent of one another. A device can be processed to dimensions as designed. In the first example, in a step illustrated in
A third example of the process of processing the gate electrode, the gate insulating film and the memory film is described hereinafter with reference to
Similarly to the second example, in the third example, the mask material 31 is deposited (see
In the case of the process according to the third example, the number of times of performing lithography is 1. As compared with the first example, the second example and a modification, the number of times of performing lithography is small. Thus, the manufacturing cost can be reduced. In the process according to the third example, patterning is performed so that the gate patterns of the memory transistors and the pass transistor are generated in the same lithography process. At that time, the element region 110 differs from the element region 111 in the height of the wafer. However, in the case of using a MONOS structure as each memory transistor, the difference in the height of a wafer is about 20 nm. Thus, patterns having dimensions as designed can be formed by selecting exposure conditions that reduce the influence of the difference in height. Alternatively, the influence of the difference in height may be reduced by preliminarily applying, before the resist is applied, an organic film, a coating-type silicon insulating film, or the like onto the mask material 31 as a resist base film.
Consequently, as compared with the third example, the number of times of performing lithographing is increased by 1. However, when the oxide film 13, the nitride film 12 and the oxide film 11 on the element region 111 are etched, the substrate surface at the element region 110 can be prevented from being etched. As compared with the modification of the second example, the number of times of performing lithographing is decreased by 1. As compared with the first example, the number of times of performing lithographing is unchanged. However, in the lithographing process illustrated in
In the above first to third examples and the modifications thereof, when the gate electrode material is processed by dry-etching using the mask material as a mask, ashing is performed to eliminate residue. At that time, if polysilicon is used as the gate electrode material, a side surface of the gate electrode is oxidized. In a state in which the side surface of the gate electrode is oxidized, if the etching of the oxide film 13, the nitride film 12 and the oxide film 11 is performed at two stages, respectively corresponding to dry-etching and wet-etching, the oxide film 13, the nitride film 12 and the oxide film 11, on the element region 111 are processed and formed into the shape of the gate electrode including the oxide film provided on the side surface. On the other hand, at the time of performing wet-etching, the oxide film provided on the side surface of the gate electrode is removed. Thus, the cross-sectional area in a (horizontal) direction parallel to the substrate of the gate electrode is size smaller than that in such a direction of the memory film. In this state, the controllability of the memory transistor on the basis of the gate voltage is low.
Even if the oxide film 13, the nitride film 12 and the oxide film 11 on the element region 111 are etched only dry-etching in a state in which the side surface of the gate electrode is oxidized, the oxidized side surface of the gate electrode is not removed. The oxidized portion of the side surface of the gate electrode does not function as an electrode. Therefore, the controllability of the memory transistor on the basis of the gate voltage is poor.
Thus, according to the first to third examples and the modifications thereof, after the gate electrode material of the memory transistor or the gate electrode materials of both of the memory transistor and the pass transistor is processed by dry-etching using the mask material as a mask, and before the oxide film 13 on the element region 111 is etched, wet-etching is performed using a solution including a hydrofluoric acid. Consequently, the oxide film 13 on the element region 111 can be etched after the oxide film on the side surface of the gate electrode is removed. Thus, the shape of the cross-section in the (horizontal) direction parallel to the substrate of the memory film is extremely close to the shape of the cross-section in the (horizontal) direction parallel to the substrate of the gate electrode.
In the above description made with reference to
However, in the case of depositing, after the STI is manufactured, the insulating film of the memory transistor, it is necessary to layout the memory transistors M11 and M21 such that the gate electrodes of the memory transistors M11 and M21 are separated from each other, as illustrated in
Usually, in the case of manufacturing chips on each of which transistors being equal or close to one another in thickness of the gate insulating film are mixedly mounted, the CMP processing of the interlayer insulating film of each transistor is performed until the gate electrode thereof is exposed. Then, the CMP processing is stopped by detecting that the gate electrode is exposed. However, according to the present embodiment, the memory film of each memory transistor differs from the gate insulating film of the pass transistor in thickness. Thus, the memory transistors and the pass transistor differ from one another in height of the surface of the gate electrode from the surface of the substrate. Accordingly, as the CMP processing proceeds, a timing at which the gate electrode of the memory transistor is exposed may differ from a timing at which the gate electrode of the pass transistor is exposed. Thus, an error may occur in the detection of the exposure of the gate electrode. Hence, the planarization processing can be performed with high controllability by performing the CMP processing in a temporally control method.
The surface of the gate electrode 14 may be planarized by CMP, as illustrated in
Even in the case of manufacturing the pass transistor PT1A as illustrated in
In a logic switch according to a second embodiment, the channel width W1 of one of the memory transistors included in a cell is larger than the channel width W2 of the other memory transistor. In the second embodiment, for example, the cell 1b illustrated in
Hereinafter, a description is made by assuming that the memory transistor M11 connected to the bit line BL1 is larger in channel width than the memory transistor M21 connected to the bit line BL2. During an operation, the bit line BL1 is connected to the ground voltage, and the bit line BL2 is connected to the power-supply voltage.
In order to prevent occurrence of a malfunction of the pass transistor during an operation of the logic switch, a voltage at the node Q is fixed to the ground voltage, or to the power-supply voltage. For example, if the signal level of, e.g., a signal input to the source or drain of the pass transistor is changed from a high level (H) to a low level (L), or if the signal level of, the signal input to the source or drain of the pass transistor is changed from L to H, a fluctuation occurs of the voltage at the node Q due to the capacitive coupling between the source/drain and the gate thereof.
Usually, even when a fluctuation of the voltage at the node Q occurs, current flows through one of the memory transistor M11 and the memory transistor M21 which is in an erased state. Thus, the voltage at the node Q returns to the ground voltage or to the power-supply voltage. A time taken to return the voltage at the node Q thereto depends upon an amount of current flowing through the memory transistor. Therefore, preferably, the channel width of the memory transistor is large. However, if the channel width is increased, a chip area increases.
Thus, an operation condition of the cell 1b in the case of causing a fluctuation of the voltage at the node Q is studied hereinafter by classifying the operation condition into four sub-conditions.
As described above, it is the third condition that may cause a malfunction of the logic switch. In the third condition, the memory transistor M11 is in an erased state, while the memory transistor M21 is in a written state. This condition corresponds to a state in which the ground voltage is supplied to the node Q via the memory transistor M11. Thus, the driving force is enhanced by increasing the channel width of the memory transistor M11, so that when the voltage at the node Q fluctuates in the third condition, a time taken to return the voltage at the node Q can be reduced.
On the other hand, in the first, second and fourth conditions, even if the voltage at the node Q fluctuates, an operation of the logic switch is not affected by the fluctuation of the voltage at the node Q. Thus, the driving force of the memory transistor M21 is permitted to be smaller than the driving force of the memory transistor M11. Hence, the channel width W2 of the memory transistor M21 is set to be smaller than the channel width W1 of the memory transistor M11. Consequently, occurrence of a malfunction of the logic switch can be prevented while increase of the chip area is suppressed.
Even if the pass transistor PT1 is a P-type transistor, the logic switch according to the present embodiment has similar advantages. In a case where the memory transistor M11 is in a written state, and where the memory transistor M21 is in an erased state, the state of the pass transistor PT1 may change to an on-state momentarily. Therefore, if the pass transistor PT1 is a P-type transistor, the channel width of the memory transistor M21 is increased, and the channel width of the memory transistor M11 is set to be smaller than the channel width of the memory transistor M21. Consequently, occurrence of a malfunction of the logic switch can be prevented while increase of the chip area is suppressed.
It is clarified by referring to computer aided design (CAD) drawings used to design a mask for lithographing that the designed values of the channel widths W1 and W2 differ from each other. Generally, after the layout of patterns is performed, correction such as optical proximity correction (OPC) is performed. However, uncorrected CAD drawings are used to check the designed values of the channels widths.
In the actual manufacture of devices, there is variation of the channel width of processed devices due to the influence of variations caused by lithography apparatuses, variations caused by resists and variations caused by base films formed on wafers. According to, e.g., International Technology Roadmap for Semiconductors (ITRS) 2009, it has been proposed, in consideration of all causes of the above variations, that the variations of pattern sizes should be suppressed so that “3*σ” (σ: a standard deviation) is within 10% of a size average. Accordingly, assuming that an average value of the channel width W1 in a chip is designated with “W1(ave)”, and that an average value of the channel width W2 in a chip is designated with “W2(ave)”, if the difference between the average values W1(ave) and W2(ave) is small, the difference therebetween may be buried in the variation thereof. However, the average value W1 is equal to or more than 10% of the average value W2(ave), the advantage in preventing a malfunction of the logic switch is promised.
The values W1(ave) and W2(ave) can be obtained by opening the manufactured chip and observing the shapes of gate electrodes with an electronic microscope or the like.
Setting the channel widths W1 and W2 at different values enables a verify operation for checking whether write operations have been performed on individual memory transistors. According to the present embodiment, verify-operations are performed by applying a predetermined verify-voltage to the word line WL1 and checking the value of the resistance between the bit lines BL1 and BL2.
When both the memory transistors M11 and M21 are in an erased state, the resistance between the bit lines BL1 and BL2 is low. On the other hand, when data is written to one of the memory transistors M11 and M21 so as to largely change a channel resistance, the resistance between the bit lines BL1 and BL2 becomes substantially equal to the channel resistance of the memory transistor to which data has been written. If the memory transistors M11 and M21 have the same structure, it cannot be determined from the resistance value between the bit lines which of the memory transistors the data has been written to. In the present embodiment, the channel widths W1 and W2 are set at different values, so that a difference in the channel resistance of the memory transistors is caused between a written state and an erased state. Therefore, it can be determined which of the memory transistors the data has been written to.
In order to achieve the verify-operations, the channel width W1 should be set to be larger than the channel width W2 in all logic cells. The variation of the pattern sizes in the manufacture of devices is considered to be within 10%. If the difference between the average values W1(ave) and W2(ave) is equal to or more than 20% in consideration of both the variation of the channel width W1 and the variation of the channel width W2, a verify-operation can be achieved.
In order to enable a verify-operation, a method of setting the gate length of the memory transistor M11 and the gate length of the memory transistor M21 at different values, and a method of setting each of the gate length and the channel width of the memory transistor M11 at a value differing from the value of an associated one of the gate length and the channel width of the memory transistor M21 are also considered. However, in this case, the threshold voltage Vth of a transistor depends upon the gate length thereof. Therefore, if the memory transistors M11 and M21 differ from each other in gate length, the threshold voltage Vth and the voltage VDsat which corresponds to occurrence of the channel pinch-off vary with individual transistors. Thus, it is necessary to change a write voltage according to which of the transistors data is written to. This means that a large number of power-supply voltages should be prepared. This results in a increase of cost. On the other hand, the method of changing the channel width between the memory transistors as described in the present embodiment affects neither the threshold voltages Vth nor the voltage VDsat. Thus, the same write voltage can be used, regardless of which of the memory transistors the data is written to.
The present embodiment has been described by assuming that the channel width of the memory transistor M11 is larger than the channel width of the memory transistor M21. However, while the logic switch operates, if the bit line BL2 is connected to the ground voltage and the bit line BL1 is connected to the power-supply voltage, the channel width of the memory transistor M21 is set to be larger than the channel width of the memory transistor M11.
The writing method and the erasure method described in the first embodiment can be used as the writing method and the erasure method in the second embodiment. In addition, the charge storage film and the block film that satisfy the requirements described in the first embodiment can be used as the charge storage film and the block film according to the second embodiment, respectively.
According to the embodiments, data can be selectively written to the memory transistor such that the threshold voltage Vth thereof is at a high level. In addition, back-tunneling can be prevented from occurring when data is erased from the memory transistor. That is, according to the embodiments, a programmable logic switch which has a small chip area and which can perform writing and erasing data without causing a malfunction can be accomplished. Further, the influence of a fluctuation of the voltage at the node Q caused when an input signal to the pass transistor changes can be reduced, and a malfunction of the programmable logic switch can be prevented, while an increase of the chip area is suppressed.
Number | Date | Country | Kind |
---|---|---|---|
2012-111974 | May 2012 | JP | national |