Programming method of nonvolatile memory cell, semiconductor memory device, and portable electronic appartaus having the semiconductor memory device

Abstract

A method for programming a nonvolatile memory cell includes the steps of: applying a first pulse to a nonvolatile memory cell to accumulate a first amount of charge which is smaller than a target amount of charge; in the case where the accumulated first amount of charge is smaller than a second amount of charge, applying a second pulse train to the nonvolatile memory cell, so that charges in an amount close to the second amount of charge are accumulated in the nonvolatile memory cell; when the nonvolatile memory cell is determined as retaining charges larger than the second amount of charge, applying a third pulse train until charges within an allowable error range of the target amount of charge is stored in the memory cell.

Description

BACKGROUND OF THE INVENTION

[0002] 1. Field of the Invention

[0003] The present invention relates to a programming method of a nonvolatile memory cell, a semiconductor memory device, a display device, and a portable electronic apparatus. More specifically, the present invention relates to a semiconductor memory device in which memory cells each including a memory functional unit having the function of retaining charges or polarization are arranged, and to a portable electronic apparatus having such a semiconductor memory device.

[0004] 2. Description of the Related Art

[0005] Conventionally, a flash memory has been typically used as a nonvolatile memory.

[0006] As shown in FIG. 22, in such a flash memory, a floating gate 902, an insulating film 907 and a word line (control gate) 903 are formed in this order on a semiconductor substrate 901 via a gate insulating film. On both sides of the floating gate 902, a source line 904 and a bit line 905 are formed in diffusion regions, thereby forming a memory cell. Around the memory cell, a device isolation region 906 is formed (see, for example, Japanese Unexamined Patent Publication No. 5-304277(1993)).

[0007] The memory cell retains data according to a charge amount in the floating gate 902. In a memory cell array constructed by arranging memory cells, by selecting a specific word line and a specific bit line and applying a predetermined voltage, an operation of rewriting/reading a desired memory cell can be performed.

[0008] In such a flash memory, when a charge amount in the floating gate changes, a drain current (Id)-gate voltage (Vg) characteristic as shown in FIG. 23 is displayed. In FIG. 23, a solid line shows the characteristics in a writing state while a dashed line shows the characteristics in an erasing state. When the amount of negative charges in the floating gate increases, the threshold increases, and an Id-Vg curve shifts almost in parallel in the direction of increasing Vg.

[0009] However, in the above-described flash memory, it has been functionally necessary to dispose the insulating film 907 for separating the floating gate 902 and the word line 903 from each other, and further, it has been difficult to reduce the thickness of the gate insulating film in order to prevent any leakage of charges from the floating gate 902. As a consequence, it has been difficult to effectively reduce the thickness of each of the insulating film 907 and the gate insulating film, thereby inhibiting the microfabrication of the memory cell.

SUMMARY OF THE INVENTION

[0010] The present invention has been achieved in consideration of such circumstances and its object is to provide a programming method of a nonvolatile memory cell of which size can be easily reduced, and a semiconductor memory device having the same.

[0011] The present invention provides a method for programming a nonvolatile memory cell comprising the steps of: applying a first pulse having a first time width at a first voltage to a nonvolatile memory cell to accumulate a first amount of charge which is smaller than a target amount of charge; testing the nonvolatile memory cell to determine whether or not the amount of charge accumulated in the memory cell is larger than a second amount of charge; in the case where the accumulated first amount of charge is smaller than the second amount of charge, applying a second pulse train to the nonvolatile memory cell, the second pulse train having pulses each having a voltage by which the second amount of charge can be accumulated with a second time width narrower than the first time width so that charges in an amount close to the second amount of charge are accumulated in the nonvolatile memory cell; after applying each pulse of the second pulse train, testing the nonvolatile memory cell to determine whether or not the memory cell retains charges larger than the second amount of charge; when the nonvolatile memory cell is determined as retaining charges larger than the second amount of charge; applying a third pulse train having pulses each having a third voltage with a third time width narrower than the second time width to the nonvolatile memory cell until charges within an allowable error range of the target amount of charge are stored in the memory cell; and wherein the nonvolatile memory cell includes a gate electrode formed on a semiconductor layer via a gate insulating film, a channel region disposed under the gate electrode, a source and a drain as diffusion regions disposed on both sides of the channel region and having a conductive type opposite to that of the channel region, and memory functional units formed on both sides of the gate electrode and having a function of retaining charges.

[0012] According to the present invention, at the time of programming a reference cell in the semiconductor memory device, charges of a first charge amount smaller than a target charge amount are accumulated at high speed, after that, a pulse train is applied to a second charge amount which is smaller than the target charge amount and is larger than the first charge amount and, further, a short pulse train is applied to a third charge amount which is within an allowable error range of the target charge amount. Consequently, the reference cell for storing a reference to be compared for reading a memory state can be programmed accurately in short time until a state where a reference is stored is obtained.

[0013] In the nonvolatile memory cell used in the semiconductor memory device of the present invention, a memory function of the memory functional unit and a transistor operation function of the gate insulating film are separated from each other. Consequently, it is easy to suppress the short channel effect by thinning the gate insulating film while making the sufficient memory function remained. Further, a value of current flowing between the diffusion regions changes by rewriting more largely than that in the case of an EEPROM. Therefore, it facilitates discrimination between the writing state and the erasing state of the semiconductor memory device.

[0014] Further, the nonvolatile memory cell in the semiconductor memory device of the present invention can be formed by a process which has very high affinity with a normal transistor forming process on the basis of the configuration. Therefore, as compared with the case of using a conventional flash memory as a nonvolatile memory cell and forming the memory cell together with a peripheral circuit which is usually made by a transistor, the number of masks and the number of processes can be dramatically reduced. Consequently, the yield in production of a chip on which both the memory cell and the peripheral circuit are formed can be improved. Because of this, the manufacturing cost is reduced, and a very-reliable, cheap semiconductor memory device can be obtained.

[0015] In the programming method according to the present invention, all of pulses may be applied to the gate electrode of the nonvolatile memory cell while maintaining the drain and the source in a predetermined state.

[0016] In the programming method according to the present invention, the first pulse may have a first voltage sufficient to make the nonvolatile memory cell reach a saturation region, each of pulses of the second pulse train may have a voltage larger than a voltage of an immediately preceding pulse, and a voltage of the third pulse train may be equal to the voltage of the final pulse of the second pulse train.

[0017] The present invention also provides a programming method of a nonvolatile memory cell including a gate electrode formed on a semiconductor layer via a gate insulating film, a channel region disposed under the gate electrode, a source and a drain as diffusion regions disposed on both sides of the channel region and having a conductive type opposite to that of the channel region, and memory functional units formed on both sides of the gate electrode and having the function of retaining charges, the method including the step of applying a first pulse having a first time width to a nonvolatile memory cell so that the nonvolatile memory cell retains a first charge amount which is smaller than a target charge amount in order to accumulate charges close to the target charge amount in the nonvolatile memory cell, applying a second pulse train of pulses each having a second charge amount with a second time width which is narrower than the first time width to the nonvolatile memory cell so that the nonvolatile memory cell retains the second charge amount smaller than the target charge amount and larger than the first charge amount, and applying a third pulse train to the nonvolatile memory cell so that the nonvolatile memory cell retains a third charge amount within an allowable error range of the target charge amount.

[0018] The programming method according to the present invention further include the step of, after applying the first pulse, testing the nonvolatile memory cell to determine whether the first charge amount exceeds the second charge amount or not, in the case where the first charge amount exceeds the second charge amount, not executing a step of applying the second pulse train, after applying each of pulses of the second pulse train, testing the nonvolatile memory cell to determine whether the nonvolatile memory cell retains the second charge amount or not, continuously applying the second pulse train until it is determined that the nonvolatile memory cell retains the second charge amount, after applying each of pulses of the third pulse train, testing the nonvolatile memory cell to determine whether the nonvolatile memory cell has retained the third charge amount or not, and continuously applying the third pulse train until it is determined that the nonvolatile memory cell has retained the third charge amount.

[0019] In the programming method according to the present invention, the voltage of each of the pulses of the second pulse train may be larger than the voltage of the immediately preceding pulse.

[0020] Further, in the programming method according to the present invention, the first pulse may have a first voltage, and the first voltage and first time width may be selected so that the nonvolatile memory cell lies in a saturation region.

[0021] The present invention also provides a semiconductor memory device including a circuit for programming a nonvolatile memory cell, wherein the programming circuit includes: means for applying a first pulse having a first time width to a nonvolatile memory cell to accumulate a first charge amount smaller than a target charge amount in a nonvolatile memory cell so that the nonvolatile memory cell retains charges close to the target charge amount; means for applying a second pulse train of pulses each having a second time width which is narrower than the first time width to the nonvolatile memory cell so that the nonvolatile memory cell retains the second charge amount which is smaller than the target charge amount and is larger than the first charge amount; and means for applying a third pulse train having a third time width which is narrower than the second time width to the nonvolatile memory device so that the nonvolatile memory cell retains a third charge amount within an allowable error range of the target charge amount, and the nonvolatile memory cell includes a gate electrode formed on a semiconductor layer via a gate insulating film, a channel region disposed under the gate electrode, a source and a drain as diffusion regions disposed on both sides of the channel region and having a conductive type opposite to that of the channel region, and memory functional units formed on both sides of the gate electrode and having the function of retaining charges.

[0022] In the semiconductor memory device according to the present invention, the programming circuit may further include: first test means for a nonvolatile memory cell, for determining, after applying the first pulse, whether the first charge amount exceeds the second charge amount or not; second test means for a nonvolatile memory cell, for determining, after applying pulses of the second pulse train, whether the nonvolatile memory cell retains the second charge amount or not; and third test means for a nonvolatile memory cell, for determining, after applying pulses of the third pulse train, whether the nonvolatile memory cell retains the third charge amount or not, and in the case where the first charge amount exceeds the second charge amount as a result of a test by the first test means, the means for applying the second pulse train is not executed, the second pulse train is continuously applied until it is determined that the second charge amount is retained as a result of a test by the second test means, and the third pulse train is continuously applied until it is determined that the third charge amount is retained as a result of the test by the third test means.

[0023] Further, in the semiconductor memory device according to the present invention, the voltage of each of the pulses of the second pulse train may be larger than the voltage of the immediately preceding pulse.

[0024] In the semiconductor memory device according to the present invention, the first pulse may have a first voltage, and the first voltage and first time width may be selected so that the nonvolatile memory cell lies in a saturation region.

[0025] Further, in the semiconductor memory device according to the present invention, the nonvolatile memory cell may have a diffusion region which is disposed so as to be offset from the gate electrode.

[0026] In the semiconductor memory device of the present invention, the nonvolatile memory cell may include a film having a surface almost parallel with a surface of the gate insulating film and having the function of retaining charges.

[0027] With the configuration, variations in the memory effect can be suppressed. Further, a large design margin for variations in the nonvolatile memory cell can be set, so that designing is facilitated.

[0028] In the semiconductor memory device of the present invention, the memory functional unit of the nonvolatile memory cell may have a charge retaining film which is an insulating film.

[0029] With the configuration, a charge leakage is prevented, and an excellent retaining characteristic is obtained. Since the nonvolatile memory cell having the excellent charge retention characteristic is used and current of a reference cell having the same cell as the nonvolatile memory cell is set accurately, reading can be performed for longer time.

[0030] In the semiconductor memory device of the present invention, the nonvolatile memory cell may include a film having a surface which extends almost parallel with a surface of the gate insulating film and having the function of retaining charges, and an insulating film for separating the film from a channel region or a semiconductor layer, and the thickness of the insulating film is larger than that of the gate insulating film and is 20 nm or less.

[0031] With the configuration, an excellent retaining characteristic is obtained. Since the nonvolatile memory cell having the excellent charge retention characteristic is used and current of a reference cell having the same cell as the nonvolatile memory cell is set accurately, reading can be performed for longer time.

[0032] The semiconductor memory device of the present invention may have a nonvolatile memory cell including a film having a surface almost parallel to a surface of a gate insulating film and having the function of retaining charges.

[0033] With the configuration, a change in the characteristics during retention is suppressed. Since the nonvolatile memory cell having an excellent charge retention characteristic is used and current of a reference cell having the same cell as the nonvolatile memory cell is set accurately, reading can be performed for longer time.

BRIEF DESCRIPTION OF THE DRAWINGS

[0034] FIG. 1 is a schematic sectional view showing a main part of a memory cell (first embodiment) in a semiconductor memory device according to the present invention;

[0035] FIG. 2.(a) and FIG. 2.(b) are schematic sectional views each showing a main part of a modification of the memory cell (first embodiment) in the semiconductor memory device according to the present invention;

[0036] FIG. 3 is a diagram for describing a writing operation of the memory cell (first embodiment) in the semiconductor memory device according to the present invention;

[0037] FIG. 4 is a diagram for describing a writing operation of the memory cell (first embodiment) in the semiconductor memory device according to the present invention;

[0038] FIG. 5 is a diagram for describing an erasing operation of the memory cell (first embodiment) in the semiconductor memory device according to the present invention;

[0039] FIG. 6 is a diagram for describing an erasing operation of the memory cell (first embodiment) in the semiconductor memory device according to the present invention;

[0040] FIG. 7 is a diagram for describing a reading operation of the memory cell (first embodiment) in the semiconductor memory device according to the present invention;

[0041] FIG. 8 is a schematic sectional view showing a main part of a memory cell (second embodiment) in the semiconductor memory device according to the present invention;

[0042] FIG. 9 is an enlarged schematic sectional view of the main part shown in FIG. 8;

[0043] FIG. 10 is an enlarged schematic sectional view of a modification of the main part shown in FIG. 8;

[0044] FIG. 11 is a graph showing electric characteristics of the memory cell (second embodiment) in the semiconductor memory device according to the present invention;

[0045] FIG. 12 is a schematic sectional view showing a main part of a modification of the memory cell (second embodiment) in the semiconductor memory device according to the present invention;

[0046] FIG. 13 is a schematic sectional view showing a main part of a memory cell (third embodiment) in the semiconductor memory device according to the present invention;

[0047] FIG. 14 is a schematic sectional view showing a main part of a memory cell (fourth embodiment) in the semiconductor memory device according to the present invention;

[0048] FIG. 15 is a schematic sectional view showing a main part of a memory cell (fifth embodiment) in the semiconductor memory device according to the present invention;

[0049] FIG. 16 is a schematic sectional view showing a main part of a memory cell (sixth embodiment) in the semiconductor memory device according to the present invention;

[0050] FIG. 17 is a schematic sectional view showing a main part of a memory cell (seventh embodiment) in the semiconductor memory device according to the present invention;

[0051] FIG. 18 is a schematic sectional view showing a main part of a memory cell (eighth embodiment) in the semiconductor memory device according to the present invention;

[0052] FIG. 19 is a graph showing electric characteristics of a memory cell (ninth embodiment) in the semiconductor memory device according to the present invention;

[0053] FIG. 20 is a schematic configuration diagram showing a liquid crystal display device (eleventh embodiment) incorporating therein the semiconductor memory device according to the present invention;

[0054] FIG. 21 is a schematic configuration diagram showing a portable electronic apparatus (twelfth embodiment) incorporating therein the semiconductor memory device according to the present invention;

[0055] FIG. 22 is a schematic sectional view showing a main part of a conventional flash memory;

[0056] FIG. 23 is a graph showing electric characteristics of a conventional flash memory;

[0057] FIG. 24 is a circuit diagram showing a semiconductor memory device (tenth embodiment) according to the present invention;

[0058] FIG. 25 is a graph showing distribution of threshold voltages in a writing state and an erasing state of a memory cell in the semiconductor memory device (tenth embodiment) according to the present invention;

[0059] FIG. 26 is a block diagram showing a circuit for reading at the time of storing four levels of information in a memory cell in the semiconductor memory device (tenth embodiment) according to the present invention;

[0060] FIG. 27 is a graph showing distribution of a threshold voltage in each memory state at the time of storing information of four values in a memory cell in the semiconductor memory device (tenth embodiment) according to the present invention;

[0061] FIG. 28 is a graph showing current-voltage characteristics of a first load transistor in the semiconductor memory device (tenth embodiment) according to the present invention; and

[0062] FIG. 29 is a flowchart showing a process for programming a reference cell in the semiconductor memory device (tenth embodiment) according to the present invention.

DETAILED DESCRIPTION OF THE INVENTION

[0063] A semiconductor memory device according to the present invention is mainly constructed by a nonvolatile memory cell, a reference cell, and a program circuit for programming the reference cell in a reference state. Herein, the program means that a desired amount of charges is accumulated in each of the nonvolatile memory cell and the reference cell. Furthermore, the program circuit is a circuit for accumulating the desired amount of charges in each of the nonvolatile memory cell and the reference cell. It is noted that the semiconductor memory device according to the present invention basically adopts a MOS circuit. It is preferable to mount all of the circuits inclusive of the MOS circuit on a single semiconductor substrate.

[0064] The nonvolatile memory cell in the semiconductor memory device according to the present invention is mainly constructed by a semiconductor layer, a gate insulating film, a gate electrode, a channel region, a diffusion region and a memory functional unit. Herein, the channel region is normally a region of the same conductive type as that of the semiconductor layer and denotes a region immediately below the gate electrode. The diffusion region denotes a region of the conductive type opposite to that of the channel region.

[0065] Concretely, the nonvolatile memory cell of the present invention may be constructed by a region of a first conductive type as a diffusion region, a region of a second conductive type as a channel region, a memory functional unit disposed across a border of the regions of the first and second conductive types, and an electrode provided via a gate insulating film. It is suitable that the nonvolatile memory cell of the present invention is constructed by a gate electrode formed on a gate insulating film, two memory functional units formed on both sides of the gate electrode, two diffusion regions disposed on the opposite sides of the gate electrode of the memory functional units, and a channel region disposed below the gate electrode. Hereinafter, the nonvolatile memory cell according to the present invention will be referred to as a sidewall memory cell.

[0066] In the semiconductor device of the present invention, the semiconductor layer is formed on the semiconductor substrate, preferably, on a well region of the first conductive type formed in the semiconductor substrate.

[0067] The semiconductor substrate is not particularly limited as long as it can be used for a semiconductor device, and an examples thereof include a bulk substrate made of an element semiconductor such as silicon, germanium or the like and a compound semiconductor such as silicon germanium, GaAs, InGaAs, ZnSe or GaN. As a substrate having a semiconductor layer on its surface, various substrates such as an SOI (Silicon on Insulator) substrate, an SOS substrate and a multilayer SOI substrate, or a glass or plastic substrate having thereon a semiconductor layer may be used. In particular, a silicon substrate and an SOI substrate having a semiconductor layer on its surface are preferable. The semiconductor substrate or semiconductor layer may be single crystal (formed by, for example, epitaxial growth), polycrystal, or amorphous although an amount of current flowing therein varies a little.

[0068] On the semiconductor layer, preferably, a device isolation region is formed. Further, a single layer or multilayer structure may be formed by a combination of devices such as a transistor, a capacitor and a resistor, a circuit formed by the devices, a semiconductor device, and an interlayer insulating film. The device isolation region can be formed by any of various device isolation films such as an LOCOS film, a trench oxide film and an STI film. The semiconductor layer may be of the P or N conductive type. In the semiconductor layer, preferably, at least one well region of the first conductive type (P or N type) is formed. As impurity concentration in the semiconductor layer and the well region, impurity concentration which is within a known range in this field can be used. In the case of using the SOI substrate as the semiconductor layer, the well region may be formed in the surface semiconductor layer and a body region may be provided below the channel region.

[0069] The gate insulating film is not particularly limited as long as it is usually used for a semiconductor device, and examples thereof include an insulating film such as a silicon oxide film or a silicon nitride film, and a high dielectric constant film such as an aluminum oxide film, a titanium oxide film, a tantalum oxide film or a hafnium oxide film, in the form of a single-layer or a laminated film. Particularly, a silicon oxide film is preferable. The gate insulating film has a thickness of, for example, about 1 to 20 nm, preferably, about 1 to 6 nm. The gate insulating film may be formed only immediately below the gate electrode or formed so as to be larger (wider) than the gate electrode.

[0070] The gate electrode is formed in a shape which is usually used for a semiconductor device or a shape having a recess in a lower end portion on the gate insulating film. The gate electrode is formed preferably in an integral form without being separated by a single-layered or multilayer conductive film. The gate electrode may be disposed in a state where it is separated by a single-layered or multilayer conductive film. The gate electrode may have a side-wall insulating film on its sidewalls. Usually, the gate electrode is not particularly limited as long as it is used for a semiconductor device, and examples of thereof include a conductive film made of polysilicon, a metal such as copper or aluminum, a high-refractory metal such as tungsten, titanium or tantalum, and a silicide with the high refractory metal, in the form of a single-layer or multilayer film. Suitable thickness of the gate electrode is, for example, about 50 to 400 nm. Below the gate electrode, a channel region is formed.

[0071] Preferably, the gate electrode is formed only on the sidewalls of the memory functional unit or does not cover the top part of the memory functional unit. By such arrangement, a contact plug can be disposed closer to the gate electrode, so that reduction in the size of the memory cell is facilitated. It is easy to manufacture the sidewall memory cell having such simple arrangement, so that the yield in production can be improved.

[0072] The memory functional unit has at least the function of retaining charges (hereinafter, described as “charge retaining function”). In other words, the memory functional unit has the function of accumulating and retaining charges, the function of trapping charges or the function of holding a charge polarization state. The function is exhibited, for example, when the memory functional unit includes a film or region having the charge retaining function. Examples of elements having the above function include: silicon nitride; silicon; a silicate glass including impurity such as phosphorus or boron; silicon carbide; alumina; a high dielectric material such as hafnium oxide, zirconium oxide or tantalum oxide; zinc oxide; ferroelectric; metals, and the like. Therefore, the memory functional unit can be formed by, for example, a single-layered or laminated structure of: an insulating film including a silicon nitride film; an insulating film having therein a conductive film or a semiconductor layer; an insulating film including at least one conductor or semiconductor dot; or an insulating film including a ferroelectric film of which inner charge is polarized by an electric field and in which the polarized state is held. Particularly, the silicon nitride film is preferable for the reason that the silicon nitride film can obtain a large hysteretic characteristic since a number of levels of trapping charges exist. In addition, the charge retention time is long and a problem of charge leakage due to occurrence of a leak path does not occur, so that the retention characteristics are good. Further, silicon nitride is a material which is used as standard in an LSI process.

[0073] By using the insulating film including a film having the charge retaining function such as a silicon nitride film as the memory functional unit, reliability of storage and retention can be increased. Since the silicon nitride film is an insulator, even in the case where a charge leak occurs in part of the silicon nitride film, the charges in the whole silicon nitride film are not lost immediately. In the case of arranging a plurality of sidewall memory cells, even when the distance between the sidewall memory cells is shortened and neighboring memory cells come into contact with each other, unlike the case where the memory functional units are made of conductors, information stored in the memory functional units is not lost. Further, a contact plug can be disposed closer to the memory functional unit. In some cases, the contact plug can be disposed so as to be overlapped with the memory functional unit. Thus, reduction in size of the memory cell is facilitated.

[0074] In order to increase the reliability of storage and retention, the film having the charge retaining function does not always have to have a film shape. Preferably, films having the charge retaining function exist discretely in an insulating film. Concretely, it is preferable that the films having the charge retaining function in the shape of dots be spread in a material which is hard to retain charges, for example, in a silicon oxide.

[0075] In the case of using a conductive film or semiconductor layer as the charge retaining film, preferably, the conductive film or semiconductor layer is disposed via an insulating film so that the charge retaining film is not in direct contact with the semiconductor layer (semiconductor substrate, well region, body region, source/drain regions or diffusion region) or a gate electrode. For example, a laminated structure of the conductive film and the insulating film, a structure in which conductive films in the form of dots are spread in the insulating film, a structure in which the conductive film is disposed in a part of a sidewall insulating film formed on sidewalls of the gate, and the like can be mentioned.

[0076] It is preferable to use the insulating film having therein the conductive film or semiconductor layer as a memory functional unit for the reason that an amount of injecting charges into the conductor or semiconductor can be freely controlled and multilevel values can be easily obtained.

[0077] Further, it is preferable to use the insulating film including at least one conductor or semiconductor dot as the memory functional unit for the reason that it becomes easier to perform writing and erasing by direct tunneling of charges, and reduction in power consumption can be achieved.

[0078] Alternatively, as a memory functional unit, a ferroelectric film such as PZT or PLZT in which the polarization direction changes according to the electric field may be used. In this case, charges are substantially generated in the surface of the ferroelectric film by the polarization and are held in that state. It is therefore preferable since the ferroelectric film can obtain a hysteresis characteristic similar to that of a film to which charges are supplied from the outside of the film having the memory function and which traps charges. In addition, it is unnecessary to inject charges from the outside of the film in order to retain charges in the ferroelectric film, and the hysteresis characteristic can be obtained only by the polarization of the charge in the film, so that writing/erasing can be performed at high speed.

[0079] As the insulating film constructing the memory functional unit, a film having a region or function of suppressing escape of charges is suitable. An example of such a film includes a silicon oxide film.

[0080] The charge retaining film included in the memory functional unit is disposed on both sides of the gate electrode directly or via an insulating film, and is disposed on the semiconductor layer (semiconductor substrate, well region, body region or source/drain region, or diffusion region) directly or via a gate insulating film. Preferably, the charge retaining film on both sides of the gate electrode is formed so as to cover all or part of the sidewalls of the gate electrode directly or via the insulating film. In an application example, in the case where the gate electrode has a recess in its lower end, the charge retaining film may be formed so as to completely or partially bury the recess directly or via an insulating film.

[0081] The diffusion regions can function as source and drain regions and have the conductive type opposite to that of the semiconductor layer or well region. In the junction, preferably, difference between impurity concentrations of the diffusion region and the semiconductor layer or well region is high for the reason that hot electrons or hot holes are generated efficiently with low voltage, and high-speed operation can be performed with lower voltage. The junction depth of the diffusion region is not particularly limited but can be appropriately adjusted in accordance with the performance or the like of a semiconductor memory device to be obtained. In the case of using an SOI substrate as a semiconductor substrate, the diffusion region may have a junction depth smaller than the thickness of the surface semiconductor layer. It is preferable that the diffusion region has junction depth almost the same as that of the surface semiconductor layer.

[0082] The diffusion region may be disposed so as to overlap with an end of the gate electrode, so as to match an end of the gate electrode, or so as to be offset from an end of the gate electrode. The case of offset is particularly preferable because easiness of inversion of the offset region below the charge retaining film largely changes in accordance with an amount of charges accumulated in the memory functional unit when voltage is applied to the gate electrode, the memory effect increases, and a short channel effect is reduced. However, when the diffusion region is offset too much, drive current between the diffusion regions (source and drain) decreases conspicuously. Therefore, it is preferable that the offset amount, that is, the distance to the diffusion area closer to one of the gate electrode ends in the gate length direction is shorter than the thickness of the charge retaining film extending in the direction parallel with the gate length direction. It is particularly important that at least a part of the film or region having the charge retaining function in the memory functional unit is overlapped with part of the diffusion region. This is because the essence of the memory cell as a component of the semiconductor memory device is to rewrite stored information by an electric field which is applied across the memory functional unit in accordance with the voltage difference between the gate electrode which exists only in the sidewall part of the memory functional unit and the diffusion region.

[0083] A part of the diffusion region may extend at a level higher than the surface of the channel region or the under face of the gate insulating film. In this case, it is suitable that, on the diffusion region formed in the semiconductor substrate, the conductive film integrated with the diffusion region is laminated. The conductive film may be made of semiconductor such as polysilicon or amorphous silicon, silicide, the above-described metals, high-refractory metals, or the like. In particular, polysilicon is preferred. Since impurity diffusion speed of polysilicon is much faster than that of the semiconductor layer, it is easy to make the junction depth of the diffusion region in the semiconductor layer shallow and to suppress the short channel effect. In this case, preferably, a part of the diffusion region is disposed so as to sandwich at least a part of the memory functional unit in cooperation with the gate electrode.

[0084] The sidewall memory cell can be formed by a normal semiconductor process, for example, a method similar to the method of forming the sidewall spacer having the single-layer or laminated structure on the sidewalls of the gate electrode. Concrete examples of the method include; a method of forming the gate electrode, after that, forming a single-layer film or laminated film including the charge retaining film such as a film having the function of retaining charges (hereinafter, described as “charge retaining film”), charge retaining film/insulating film, insulating film/charge retaining film, or insulating film/charge retaining film/insulating film, and etching back the formed film under suitable conditions so as to leave the films in a sidewall spacer shape; a method of forming an insulating film or charge retaining film, etching back the film under suitable conditions so as to leave the film in the sidewall spacer shape, further forming the charge retaining film or insulating film, and similarly etching back the film so as to leave the film in the sidewall spacer shape; a method of applying or depositing an insulating film material in which particles made of a charge retaining material are spread on the semiconductor layer including the gate electrode and etching back the material under suitable conditions so as to leave the insulating film material in a sidewall spacer shape; and a method of forming a gate electrode, after that, forming the single-layer film or laminated film, and patterning the film with a mask. According to another method, before the gate electrode is formed, charge retaining film, charge retaining film/insulating film, insulating film/charge retaining film, insulating film/charge retaining film/insulating film, or the like is formed. An opening is formed in a region which becomes the channel region of the films, a gate electrode material film is formed on the entire surface of the opening, and the gate electrode material film is patterned in a shape including the opening and larger than the opening, thereby forming the gate electrode and the memory functional unit.

[0085] One example of a method for forming the sidewall memory cell according to the present invention will now be described. First, the gate insulating film and the gate electrode are formed on the semiconductor substrate in accordance with known procedures. Subsequently, a silicon oxide film having a thickness of 0.8 to 20 nm, more preferably 3 to 10 nm is formed by thermal oxidation or deposited by CVD (Chemical Vapor Deposition) over the entire semiconductor substrate. Next, a silicon nitride film having a thickness of 2 to 15 nm, more preferably 3 to 10 nm is deposited by the CVD over the entire silicon oxide film. Moreover, another silicon oxide film having a thickness of 20 to 70 nm is deposited by the CVD over the entire silicon nitride film.

[0086] Subsequently, the silicon oxide film/silicon nitride film/silicon oxide film are etched back by anisotropic etching, thereby forming the memory functional unit optimum for storing data on the sidewall of the gate electrode in the form of a sidewall spacer.

[0087] Thereafter, ions are injected while using the gate electrode and the memory functional unit in the form of the sidewall spacer as masks, thereby forming a diffusion layer region (source/drain region). After that, a silicide process or an upper wiring process may be performed in accordance with known procedures.

[0088] In the case of constructing the memory cell array by arranging sidewall memory cells, the best mode of the sidewall memory cell satisfies all of the requirements: for example, (1) the gate electrodes of a plurality of sidewall memory cells are integrated and have the function of a word line; (2) the memory functional units are formed on both sides of the word line; (3) an insulator, particularly, a silicon nitride film retains charges in the memory functional unit; (4) the memory functional unit is constructed by an ONO (Oxide Nitride Oxide) film and the silicon nitride film has a surface almost parallel with the surface of the gate insulating film; (5) a silicon nitride film in the memory functional unit is isolated from a word line and a channel region via a silicon oxide film; (6) the silicon nitride film and a diffusion region in the memory functional unit are overlapped; (7) the thickness of the insulating film separating the silicon nitride film having the surface which is almost parallel with the surface of the gate insulating film from the channel region or semiconductor layer and the thickness of the gate insulating film are different from each other; (8) an operation of writing/erasing one sidewall memory cell is performed by a single word line; (9) there is no electrode (word line) having the function of assisting the writing/erasing operation on the memory functional unit; and (10) in a portion in contact with the diffusion region immediately below the memory functional unit, a region of high concentration of impurity whose conductive type is opposite to that of the diffusion region is provided. It is sufficient for the memory cell to satisfy even one of the requirements.

[0089] A particularly preferable combination of the requirements is, for example, (3) an insulator, particularly, a silicon nitride film retains charges in the memory functional unit, (6) the insulating film (silicon nitride film) and the diffusion region in the memory functional unit are overlapped, and (9) there is no electrode (word line) having the function of assisting the writing/erasing operation on the memory functional unit.

[0090] In the case where the sidewall memory cell satisfies the requirements (3) and (9), it is very useful for the following reasons. First, the bit line contact can be disposed closer to the memory functional unit on the word line sidewall or even when the distance between sidewall memory cells is shortened, a plurality of memory functional units do not interfere with each other, and stored information can be held. Therefore, reduction in size of the memory cell is facilitated. In the case where the charge retaining region in the memory functional unit is made of a conductor, as the distance between sidewall memory cells decreases, interference occurs between the charge retaining regions due to capacitive coupling, so that stored information cannot be held.

[0091] In the case where the charge retaining region in the memory functional unit is made of an insulator (for example, a silicon nitride film), it becomes unnecessary to make the memory functional unit independent for each sidewall memory cell. For example, the memory functional units formed on both sides of a single word line shared by a plurality of sidewall memory cells do not have to be isolated for each sidewall memory cell. The memory functional units formed on both sides of one word line can be shared by a plurality of sidewall memory cells sharing the word line. Consequently, a photo etching process for isolating the memory functional unit becomes unnecessary, and the manufacturing process is simplified. Further, a margin for positioning in the photolithography process and a margin for film reduction by etching become unnecessary, so that the margin between neighboring sidewall memory cells can be reduced. Therefore, as compared with the case where the charge retaining region in the memory functional unit is made of a conductor (for example, polysilicon film), even when the memory functional unit is formed at the same microfabrication level, a sidewall memory cell occupied area can be reduced. In the case where the charge retaining region in the memory functional unit is made of a conductor, the photo etching process for isolating the memory functional unit for each sidewall memory cell is necessary, and a margin for positioning in the photolithography process and a margin for film reduction by etching are necessary.

[0092] Moreover, since the electrode having the function of assisting the writing and erasing operations does not exist on the memory functional unit and the device structure is simple, the number of processes decreases, so that the yield in production can be increased. Therefore, it facilitates formation with a transistor as a component of a logic circuit or an analog circuit, and a cheap semiconductor memory device can be obtained.

[0093] The present invention is more useful in the case where not only the requirements (3) and (9) but also the requirement (6) are satisfied.

[0094] Specifically, by overlapping the charge retaining region in the memory functional unit and the diffusion region, writing and erasing can be performed with a very low voltage. Concretely, with a low voltage of 5 V or less, the writing and erasing operations can be performed. The action is a very large effect also from the viewpoint of circuit designing. Since it is unnecessary to generate a high voltage in a chip unlike a flash memory, a charge pumping circuit requiring a large occupation area can be omitted or its scale can be reduced. Particularly, when a memory of small-scale capacity is provided for adjustment in a logic LSI, as for an occupied area in a memory part, an occupation area of peripheral circuits for driving a sidewall memory cell is dominant more than that of a sidewall memory cell. Consequently, omission or down sizing of the voltage boosting circuit for a sidewall memory cell is most effective to reduce the chip size.

[0095] On the other hand, in the case where the requirement (3) is not satisfied, that is, in the case where a conductor retains charges in the memory functional unit, even when the requirement (6) is not satisfied, specifically, even when the conductor in the memory functional unit and the diffusion region do not overlap with each other, writing operation can be performed. This is because that the conductor in the memory functional unit assists writing operation by capacitive coupling with the gate electrode.

[0096] In the case where the requirement (9) is not satisfied, specifically, in the case where the electrode having the function of assisting the writing and erasing operations exists on the memory functional unit, even when the requirement (6) is not satisfied, specifically, even when the insulator in the memory functional unit and the diffusion region do not overlap with each other, writing operation can be performed.

[0097] In the semiconductor memory device of the present invention, a transistor may be connected in series with one of or both sides of a sidewall memory cell, or the sidewall memory cell may be mounted on the same chip with a logic transistor. In such a case, the semiconductor device of the present invention, particularly, the sidewall memory cell can be formed by a process having high compatibility with a process of forming a normal standard transistor such as a transistor or a logic transistor, they can be formed simultaneously. Therefore, a process of forming both the sidewall memory cell and a transistor or a logic transistor is very simple and, as a result, a cheap embedded device can be obtained.

[0098] In the semiconductor memory device of the present invention, the sidewall memory cell can store information of two or more values in one memory functional unit. Thus, the sidewall memory cell can function as a memory cell for storing information of four or more values. The sidewall memory cell may store binary data only. The sidewall memory cell is also allowed to function as a memory cell having the functions of both a selection transistor and a memory transistor by a variable resistance effect of the memory functional unit.

[0099] The semiconductor memory device of the present invention can be widely applied by being combined with a logic device, a logic circuit or the like to: a data processing system such as a personal computer, a note-sized computer, a laptop computer, a personal assistant/transmitter, a mini computer, a workstation, a main frame, a multiprocessor/computer, a computer system of any other type, or the like; an electronic part as a component of the data processing system, such as a CPU, a memory or a data memory device; a communication apparatus such as a telephone, a PHS, a modem or a router; an image display apparatus such as a display panel or a projector; an office apparatus such as a printer, a scanner or a copier; an image pickup apparatus such as a video camera or a digital camera; an entertainment apparatus such as a game machine or a music player; an information apparatus such as a portable information terminal, a watch or an electronic dictionary; a vehicle-mounted apparatus such as a car navigation system or a car audio system; an AV apparatus for recording/reproducing information such as a motion picture, a still picture or music; an appliance such as a washing machine, a microwave, a refrigerator, a rice cooker, a dish washer, a vacuum cleaner or an air conditioner; a health managing apparatus such as a massage device, a bathroom scale or a manometer; and a portable memory device such as an IC card or a memory card. Particularly, it is effective to apply the semiconductor memory device to portable electronic apparatuses such as portable telephone, portable information terminal, IC card, memory card, portable computer, portable game machine, digital camera, portable motion picture player, portable music player, electronic dictionary and watch. The semiconductor memory device of the present invention may be provided as at least a part of a control circuit or a data storing circuit of an electronic device or, as necessary, detachably assembled.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0100] Hereinafter, embodiments of the semiconductor memory device, the display device and the portable electronic apparatus of the present invention will be described in detail with reference to the drawings.

[0101] First Embodiment

[0102] A semiconductor memory device of a first embodiment has a sidewall memory cell 1 as shown in FIG. 1.

[0103] The sidewall memory cell 1 has a gate electrode 104 formed on a P-type well region 102 formed on the surface of a semiconductor substrate 101 via a gate insulating film 103. On the top face and side faces of the gate electrode 104, a silicon nitride film 109 having a trap level of retaining charges and serving as a charge retaining film is disposed. In the silicon nitride film 109, parts of both sidewalls of the gate electrode 104 serve as memory functional units 105a and 105b for actually retaining charges. The memory functional unit refers to a part in which charges are actually accumulated by rewriting operation in the memory functional unit or the charge retaining film. In the P-type well region 102 on both sides of the gate electrode 104, N-type diffusion regions 107a and 107b functioning as a source region and a drain region, respectively, are formed. Each of the diffusion regions 107a and 107b has an offset structure. Specifically, the diffusion regions 107a and 107b do not reach a region 121 below the gate electrode 104, and offset regions 120 below the charge retaining film construct part of the channel region.

[0104] The memory functional units 105a and 105b for substantially retaining charges are the parts on both sidewalls of the gate electrode 104. It is therefore sufficient that the silicon nitride film 109 is formed only in regions corresponding to the parts (see FIG. 2.(a)). Each of the memory functional units 105a and 105b may have a structure in which fine particles 111 each made of a conductor or semiconductor and having a nanometer size are distributed like discrete points in an insulating film 112 (see FIG. 2.(b)). When the fine particle 111 has a size less than 1 nm, a quantum effect is too large, so that it becomes hard for charges to go (tunnel) to the dots. When the size exceeds 10nm, a conspicuous quantum effect does not appear at room temperature. Therefore, the diameter of the fine particle 111 is preferably in a range from 1 nm to 10 nm. The silicon nitride film 109 serving as a charge retaining film may be formed in a sidewall spacer shape on a side face of the gate electrode (see FIG. 3).

[0105] The principle of the writing operation of the sidewall memory cell will be described with reference to FIGS. 3 and 4. The case where whole memory functional units 131a and 131b have the function of retaining charges will be described. “Writing” denotes herein injection of electrons into the memory functional units 131a and 131b when the sidewall memory cell is of the N channel type. Hereinafter, on assumption that the sidewall memory cell is of the N channel type, description will be given.

[0106] In order to inject electrons (write) the second memory functional unit 131b, as shown in FIG. 3, the first diffusion region 107a of the N type is set as a source electrode, and the second diffusion region 107b of the N type is set as a drain electrode. For example, 0 V is applied to the first diffusion region 107a and the P-type well region 102, +5 V is applied to the second diffusion region 107b, and +5 V is applied to the gate electrode 104. Under such voltage parameters, an inversion layer 226 extends from the first diffusion region 107a (source electrode) but does not reach the second diffusion region 107b (drain electrode), and a pinch off point occurs. Electrons are accelerated from the pinch-off point to the second diffusion region 107b (drain electrode) by a high electric field, and become so-called hot electrons (high-energy conduction electrons). By injection of the hot electrons into the second memory functional unit 131b, writing is performed. Since hot electrons are not generated in the vicinity of the first memory functional unit 131a, writing is not performed.

[0107] On the other hand, in order to inject electrons (write) into the first memory functional unit 131a, as shown in FIG. 4, the second diffusion region 107a is set as the source electrode, and the first diffusion region 107a is set as the drain electrode. For example, 0 V is applied to the second diffusion region 107b and the P-type well region 102, +5 V is applied to the first diffusion region 107a, and +5 V is applied to the gate electrode 104. By interchanging the source and drain regions so as to be different from the case of injecting electrons into the second memory functional unit 131b, electrons are injected into the first memory functional unit 131a and writing can be performed.

[0108] The principle of erasing operation of the sidewall memory cell will now be described with reference to FIGS. 5 and 6.

[0109] In a first method of erasing information stored in the first memory functional unit 131a, by applying positive voltage (for example, +5 V) to the first diffusion region 107a and applying 0 V to the P-type well region 102 as shown in FIG. 5, the PN junction between the first diffusion region 107a and the P-type well region 102 is reverse-biased and, further, negative voltage (for example, −5 V) is applied to the gate electrode 104. At this time, in the vicinity of the gate electrode 104 in the PN junction, due to the influence of the gate electrode to which the negative voltage is applied, particularly, gradient of potential becomes sharp. Consequently, hot holes (positive holes of high energy) are generated on the side of the P-type well region 102 of the PN junction by interband tunneling. The hot holes are attracted toward the gate electrode 104 having a negative potential and, as a result, the holes are injected to the first memory functional unit 131a. In such a manner, information in the first memory functional unit 131a is erased. At this time, to the second diffusion region 107b, it is sufficient to apply 0 V.

[0110] In the case of erasing information stored in the second memory functional unit 131b, the above-described operation is performed while interchanging the potential of the first diffusion region and that of the second diffusion region.

[0111] In a second method of erasing information stored in the first memory functional unit 131a, as shown in FIG. 6, positive voltage (for example, +4 V) is applied to the first diffusion region 107a, 0 V is applied to the second diffusion region 107b, negative voltage (for example, −4 V) is applied to the gate electrode 104, and positive voltage (for example, +0.8 V) is applied to the P-type well region 102. At this time, forward voltage is applied between the P-type well region 102 and the second diffusion region 107b, and electrons are injected to the P-type well region 102. The injected electrons are diffused to the PN junction between the P-type well region 102 and the first diffusion region 107a, where the electrons are accelerated by a strong electric field, thereby becoming hot electrons. By the hot electrons, an electron-hole pair is generated in the PN junction. Specifically, by applying forward voltage between the P-type well region 102 and the second diffusion region 107b, electrons injected in the P-type well region 102 become a trigger, and hot holes are generated in the PN junction positioned on the opposite side. The hot holes generated in the PN junction are attracted toward the gate electrode 104 having the negative potential and, as a result, positive holes are injected into the first memory functional unit 131a.

[0112] According to the method, also in the case where only voltage insufficient to generate hot holes by interband tunneling is applied to the PN junction between the P-type well region and the first diffusion region 107a, electrons injected from the second diffusion region 107b become a trigger to generate an electron-positive hole pair in the PN junction, thereby enabling hot holes to be generated. Therefore, voltage in the erasing operation can be decreased. Particularly, in the case where the offset region 120 (see FIG. 1) exists, an effect that the gradient of potential in the PN junction becomes sharp by the gate electrode to which the negative potential is applied is low. Consequently, although it is difficult to generate hot holes by interband tunneling, by the second method, the disadvantage is overcome and the erasing operation can realized with low voltage.

[0113] In the case of erasing information stored in the first memory functional unit 131a, +5 V has to be applied to the first diffusion region 107a in the first erasing method whereas +4 V is sufficient in the second erasing method. As described above, according to the second method, the voltage at the time of erasing can be decreased, so that power consumption can be reduced and deterioration of the sidewall memory cell due to hot carriers can be suppressed.

[0114] In any of the erasing methods, over-erasure does not occur easily in the sidewall memory cell. The over-erasure herein denotes a phenomenon that as the amount of positive holes accumulated in the memory functional unit increases, the threshold decreases without saturation. The over-erasure is a big issue in an EEPROM typified by a flash memory. Particularly, in the case where the threshold becomes negative, critical malfunctioning that selection of a sidewall memory cell becomes impossible occurs. On the other hand, in the sidewall memory cell in the semiconductor memory device of the present invention, also in the case where a large amount of positive holes are accumulated in the memory functional unit, only electrons are induced below the memory functional unit but an influence is hardly exerted to the potential in the channel region below the gate insulating film. Since the threshold at the time of erasing is determined by the potential below the gate insulating film, occurrence of over-erasure is suppressed.

[0115] Further, the principle of reading operation of the sidewall memory cell will be described with reference to FIG. 7.

[0116] In the case of reading information stored in the first memory functional unit 131a, the first diffusion region 107a is set as a source electrode, the second diffusion region 107b is set as a drain electrode, and the transistor is allowed to operate. For example, 0 V is applied to the first diffusion region 107a and the P-type well region 102, +1.8 V is applied to the second diffusion region 107b, and +2 V is applied to the gate electrode 104. In the case where electrons are not accumulated in the first memory functional unit 131a at this time, drain current is apt to flow. On the other hand, in the case where electrons are accumulated in the first memory functional unit 131a, an inversion layer is not easily formed in the vicinity of the first memory functional unit 131a, so that the drain current is not apt to flow. Therefore, by detecting the drain current, information stored in the first memory functional unit 131a can be read. In the case of applying a voltage so as to perform the pinch-off operation, thereby reading information, it is possible to determine with higher accuracy the state of charge accumulation in the first memory functional unit 131a without influence of the presence/absence of charge accumulation in the second memory functional unit 131b.

[0117] In the case of reading information stored in the second memory functional unit 131b, the second diffusion region 107b is set as a source electrode, the first diffusion region 107a is set as a drain electrode, and the transistor is operated. It is sufficient to apply, for example, 0V to the second diffusion region 107b and the P-type well region 102, +1.8 V to the first diffusion region 107a, and +2 V to the gate electrode 104. By interchanging the source and drain regions of the case of reading information stored in the first memory functional unit 131a, information stored in the second memory functional unit 131b can be read.

[0118] In the case where a channel region (offset region 120) which is not covered with the gate electrode 104 remains, in the channel region which is not covered with the gate electrode 104, an inversion layer is dissipated or formed according to the presence/absence of excessive charges in the memory functional units 131a and 131b and, as a result, large hysteresis (change in the threshold) is obtained. However, when the offset region 120 is too wide, the drain current largely decreases and reading speed becomes much slower. Therefore, it is preferable to determine the width of the offset region 120 so as to obtain sufficient hysteresis and reading speed.

[0119] Also in the case where the diffusion regions 107a and 107b reach ends of the gate electrode 104, that is, the diffusion regions 107a and 107b overlap with the gate electrode 104, the threshold of the transistor hardly changes by the writing operation. However, parasitic resistance at the source/drain ends largely changes, and the drain current largely decreases (by equal to or more than one digit). Therefore, reading can be performed by detecting the drain current, and the function as a memory can be obtained. In the case where a larger memory hysteresis effect is necessary, it is preferable that the diffusion regions 107a and 107b and the gate electrode 104 are not overlapped (offset region 120 exists).

[0120] By the above operating method, two bits can be written/erased selectively per one transistor. By connecting a word line WL to the gate electrode 104 of the memory cell, connecting a first bit line BL1 to the first diffusion region 107a, connecting a second bit line BL2 to the second diffusion region 107b, and arranging memory cells, a memory cell array can be constructed.

[0121] In the above-described operating method, by interchanging the source electrode and the drain electrode, writing and erasing of two bits per one transistor are performed. Alternately, by fixing the source electrode and the drain electrode, the transistor may operate as a 1-bit memory. In this case, common fixed voltage can be applied to one of the source and drain regions, so that the number of bit lines connected to the source/drain regions can be reduced to the half.

[0122] As obvious from the above description, in the sidewall memory cell in the semiconductor memory device of the present invention, the memory functional unit is formed independently of the gate insulating film, and is formed on both sides of the gate electrode, so that 2-bit operation is possible. Since each memory functional unit is isolated by the gate electrode, interference at the time of rewriting is effectively suppressed. Further, since the gate insulating film is isolated from the memory functional unit, it can be formed thinly and a short channel effect can be suppressed. Therefore, reduction in size of the memory cell and, accordingly, the semiconductor memory device can be achieved easily.

[0123] Second Embodiment

[0124] A sidewall memory cell in a semiconductor memory device according to a second embodiment has a configuration substantially similar to that of the sidewall memory cell 1 of FIG. 1 except that, as shown in FIG. 8, each of memory functional units 261 and 262 is constructed by a charge retaining region (which is a charge accumulating region and may be a film having the function of retaining charges) and a region for suppressing escape of charges (or a film having the function of suppressing escape of charges).

[0125] From the viewpoint of improving a memory retention characteristic, preferably, the memory functional unit includes a charge retaining film having the function of retaining charges and an insulating film. In the second embodiment, a silicon nitride film 242 having a level of trapping charges is used as the charge retaining film, and silicon oxide films 241 and 243 having the function of preventing dissipation of charges accumulated in the charge retaining are used as insulating films. The memory functional unit includes the charge retaining film and the insulating films, thereby preventing dissipation of charges, and the retention characteristic can be improved. As compared with the case where the memory functional unit is constructed only by the charge retaining film, the volume of the charge retaining film is appropriately reduced, movement of charges in the charge retaining film is regulated, and occurrence of a characteristic change due to charge movement during retention of information is suppressed. Further, by employing the structure in which the silicon nitride film 242 is sandwiched by the silicon oxide films 241 and 243, charge injecting efficiency at the time of rewriting operation becomes high, so that higher-speed operation can be performed. In the sidewall memory cell, the silicon nitride film 242 may be replaced with a ferroelectric.

[0126] The regions for retaining charges (silicon nitride films 242) in the memory functional units 261 and 262 overlap with diffusion regions 212 and 213. The overlap denotes herein that at least a part of the region for retaining charges (silicon nitride film 242) exists over at least a part of the diffusion regions 212 and 213. A reference numeral 211 denotes a semiconductor substrate, a reference numeral 214 denotes a gate insulating film, a reference numeral 217 denotes a gate electrode, and a reference numeral 271 indicates an offset region between the gate electrode 217 and the diffusion regions 212 and 213. Although not shown, the surface of the semiconductor substrate 211 under the gate insulating film 214 serves as a channel region.

[0127] An effect obtained when the silicon nitride films 242 as regions for retaining charges in the memory functional units 261 and 262 overlap with the diffusion regions 212 and 213 will be described.

[0128] As shown in FIG. 9, in an area around the memory functional unit 262, when an offset amount between the gate electrode 217 and the diffusion region 213 is W1 and the width of the memory functional unit 262 in a cross section in the channel length direction of the gate electrode is W2, the overlap amount between the memory functional unit 262 and the diffusion region 213 is expressed as W2−W1. It is important herein that the silicon oxide film 242 in the memory functional unit 262 overlaps with the diffusion region 213, that is, the relation of W2>W1 is satisfied.

[0129] In FIG. 9, an end on the side apart from the gate electrode 217 of the silicon nitride film 242 in the memory functional unit 262 matches with the end of the memory functional unit 262 on the side apart from the gate electrode 217, so that the width of the memory functional unit 262 is defined as W2.

[0130] As shown in FIG. 10, when the end on the side apart from the gate electrode of a silicon nitride film 242a in a memory functional unit 262a does not match with the end of the memory functional unit 262a on the side apart from the gate electrode, W2 may be defined as a distance from the gate electrode end to an end on the side apart from the gate electrode of the silicon nitride film 242a.

[0131] FIG. 11 shows drain current Id when the width W2 of the memory functional unit 262 is fixed to 100 nm and the offset amount W1 is changed in the structure of the sidewall memory cell of FIG. 9. Herein, the drain current was obtained by device simulation on assumption that the memory functional unit 262 is in erasing state (holes are accumulated), and the diffusion regions 212 and 213 serve as the source electrode and the drain electrode, respectively.

[0132] As obvious from FIG. 11, in the range where W1 is 100 nm or more (that is, the silicon nitride film 242 and the diffusion region 213 do not overlap with each other), the drain current sharply decreases. Since the drain current value is almost proportional to the reading operation speed, the performance of the memory sharply deteriorates with W1 of 100 nm or more. On the other hand, in the range where the silicon nitride film 242 and the diffusion region 213 overlap with each other, decrease in the drain current is gentle. Therefore, in the case of considering also variations in mass production, if at least a part of the silicon nitride film 242 as the film having the function of retaining charges does not overlap with the source and drain regions, it is difficult to obtain the memory function in reality.

[0133] On the basis of the result of the device simulation, by fixing W2 to 100 nm and setting W1 to 60 nm and 100 nm as design values, sidewall memory cell arrays were produced. In the case where W1 is 60 nm, the silicon nitride film 242 and the diffusion regions 212 and 213 overlap with each other by 40 nm as a design value. In the case where W1 is 100 nm, there is no overlap as a design value. Reading time of the sidewall memory cell arrays was measured and worst cases considering variations were compared with each other. In the where W1 is set to 60 nm as a design value, read access time is 100 times as fast as that of the other case. In practice, the read access time is preferably 100 n/sec or less per one bit. When W1=W2, the condition cannot be achieved. In the case of considering manufacture variations as well, it is more preferable that (W2−W1)>10 nm.

[0134] To read information stored in the memory functional unit 261 (region 281), in a manner similar to the first embodiment, it is preferable to set the diffusion region 212 as a source electrode, set the diffusion region 213 as a drain region, and form a pinch-off point on the side closer to the drain region in the channel region. Specifically, at the time of reading information stored in one of the two memory functional units, it is preferable to form the pinch-off point in a region closer to the other memory functional unit, in the channel region. With the arrangement, irrespective of a storage state of the memory functional unit 262, information stored in the memory functional unit 261 can be detected with high sensitivity, and it is a large factor to achieve 2-bit operation.

[0135] On the other hand, in the case of storing information only one of two memory functional units or in the case of using the two memory functional units in the same storage state, it is not always necessary to form the pinch-off point at the time of reading.

[0136] Although not shown in FIG. 8, it is preferable to form a well region (P-type well in the case of the N channel device) in the surface of the semiconductor substrate 211. By forming the well region, it becomes easy to control the other electric characteristics (withstand voltage, junction capacitance and short-channel effect) while setting the impurity concentration in the channel region optimum to the memory operations (rewriting operation and reading operation).

[0137] The memory functional unit preferably includes the charge retaining film disposed almost in parallel with the gate insulating film surface. In other words, it is preferable that the level of the top face of the charge retaining film in the memory functional unit is positioned parallel to the level of the top face of the gate insulating film. Concretely, as shown in FIG. 12, the silicon nitride film 242a as a charge retaining film of the memory functional unit 262 has a surface almost parallel with the surface of the gate insulating film 214. In other words, it is preferable that the silicon nitride film 242a is formed at a level parallel to the level corresponding to the surface of the gate insulating film 214.

[0138] By the existence of the silicon nitride film 242a almost parallel to the surface of the gate insulating film 214 in the memory functional unit 262, formation easiness of the inversion layer in the offset region 271 can be effectively controlled in accordance with an amount of charges accumulated in the silicon nitride film 242a. Thus, the memory effect can be increased. By forming the silicon nitride film 242a almost in parallel with the surface of the gate insulating film 214, even in the case where the offset amount (W1) varies, a change in the memory effect can be maintained relatively small, and variations of the memory effect can be suppressed. Moreover, movement of the charges upward in the silicon nitride film 242a is suppressed, and occurrence of a characteristic change due to the charge movement during retention of information can be suppressed.

[0139] Preferably, the memory functional unit 262 includes an insulating film (for example, portion on the offset region 271 in the silicon oxide film 244) for separating the silicon nitride film 242a which is almost parallel to the surface of the gate insulating film 214 and the channel region (or well region). By the insulating film, dissipation of the charges accumulated in the charge retaining film is suppressed and a sidewall memory cell having a better retention characteristic can be obtained.

[0140] By controlling the thickness of the silicon nitride film 242a and controlling the thickness of the insulating film below the silicon nitride film 242a (portion on the offset region 271 in the silicon oxide film 244) to be constant, the distance from the surface of the semiconductor substrate to charges accumulated in the charge retaining film can be maintained almost constant. To be specific, the distance from the surface of the semiconductor substrate to the charges accumulated in the charge retaining film can be controlled in a range from the minimum thickness value of the insulating film under the silicon nitride film 242a to the sum of the maximum thickness value of the insulating film under the silicon nitride film 242a and the maximum thickness value of the silicon nitride film 242a. Consequently, density of electric lines of force generated by the charges accumulated in the silicon nitride film 242a can be almost controlled, and variations in the memory effect of the sidewall memory cell can be reduced very much.

[0141] Third Embodiment

[0142] The memory functional unit 262 in a semiconductor memory device of a third embodiment has a shape in which the silicon nitride film 242 as a charge retaining film has almost uniform thickness and is disposed almost in parallel with the surface of the gate insulating film 214 as shown in FIG. 13 (region 281) and, further, almost in parallel with a side face of the gate electrode 217 (region 282).

[0143] In the case where positive voltage is applied to the gate electrode 217, an electric line 283 of force in the memory functional unit 262 passes the silicon nitride film 242 twice (regions 282 and 281) as shown by an arrow. When negative voltage is applied to the gate electrode 217, the direction of the electric line of force becomes opposite. Herein, the dielectric constant of the silicon nitride film 242 is about 6, and that of silicon oxide films 241 and 243 is about 4. Therefore, effective dielectric constant of the memory functional unit 262 in the direction of the electric line 283 of force is higher and the potential difference at both ends of the electric line of force can be reduced more as compared with the case where only the region 281 of the charge retaining film exists. In other words, a large part of the voltage applied to the gate electrode 217 is used to enhance the electric field in the offset region 271.

[0144] The reason why charges are injected to the silicon nitride film 242 in the rewriting operation is because generated charges are attracted by the electric field in the offset region 271. Therefore, by including the charge retaining film shown by the arrow 282, charges injected into the memory functional unit 262 increase in the rewriting operation, and the rewriting speed increases.

[0145] In the case where the portion of the silicon oxide film 243 is also the silicon nitride film, that is, in the case where the level of the charge retaining film is not parallel with the level corresponding to the surface of the gate insulating film 214, upward movement of charges in the silicon nitride film becomes conspicuous, and the retention characteristic deteriorates.

[0146] More preferably, in place of the silicon nitride film, the charge retaining film is made of a high dielectric such as hafnium oxide having a very high dielectric constant.

[0147] It is preferable that the memory functional unit further includes an insulating film (portion on the offset region 271 in the silicon oxide film 241) for separating the charge retaining film almost parallel to the surface of the gate insulating film and the channel region (or well region). By the insulating film, dissipation of charges accumulated in the charge retaining film is suppressed, and the retention characteristic can be further improved.

[0148] Preferably, the memory functional unit further includes an insulating film (portion in contact with the gate electrode 217 in the silicon oxide film 241) for separating the gate electrode and the charge retaining film extended almost parallel with the side face of the gate electrode. The insulating film prevents injection of charges from the gate electrode into the charge retaining film and accordingly prevents a change in the electric characteristics. Thus, the reliability of the sidewall memory cell can be improved.

[0149] Further, in a manner similar to the second embodiment, it is preferable to control the thickness of the insulating film under the silicon nitride film 242 (portion on the offset region 271 in the silicon oxide film 241) to be constant and to control the thickness of the insulating film on the side face of the gate electrode (portion in contact with the gate electrode 217 in the silicon oxide film 241) to be constant. Consequently, the density of the electric lines of force generated by the charges accumulated in the silicon nitride film 242 can be almost controlled, and charge leak can be prevented.

[0150] Fourth Embodiment

[0151] In a fourth embodiment, optimization of the gate electrode, the memory functional unit, and the distance between the source and drain regions of a sidewall memory cell in a semiconductor memory device will be described.

[0152] As shown in FIG. 14, a reference character A denotes length of the gate electrode in a cut surface in the channel length direction, a reference character B denotes the distance between the source and drain regions (channel length), and a reference character C denotes the distance from the end of one of memory functional units to the end of the other memory functional unit, that is, the distance between the end (on the side far from the gate electrode) of a film having the function of retaining charges in one of memory functional units to the end (on the side apart from the gate electrode) of a film having the function of retaining charges in the other memory functional unit in a cut surface in the channel length direction.

[0153] In such a sidewall memory cell, B<C is preferable. By satisfying such a relation, the offset regions 271 exist between the portion under the gate electrode 217 in the channel region and the diffusion regions 212 and 213. Consequently, easiness of inversion effectively fluctuates in the whole offset regions 271 by charges accumulated in the memory functional units 261 and 262 (silicon nitride films 242). Therefore, the memory effect increases and, particularly, higher-speed reading operation is realized.

[0154] In the case where the gate electrode 217 and the diffusion regions 212 and 213 are offset from each other, that is, in the case where the relation of A<B is satisfied, easiness of inversion in the offset region when voltage is applied to the gate electrode largely varies according to an amount of charges accumulated in the memory functional unit, so that the memory effect increases, and the short channel effect can be reduced.

[0155] However, as long as the memory effect appears, the offset region 271 does not always have to exist. Also in the case where the offset region 271 does not exist, if the impurity concentration in the diffusion regions 212 and 213 is sufficiently low, the memory effect can be exhibited in the memory functional units 261 and 262 (silicon nitride films 242).

[0156] Therefore, A<B<C is the most preferable.

[0157] Fifth Embodiment

[0158] A sidewall memory cell of a semiconductor memory device in a fifth embodiment has a substantially similar configuration to that of the second embodiment except that an SOI substrate is used as the semiconductor substrate in the second embodiment as shown in FIG. 15.

[0159] In the sidewall memory cell, a buried oxide film 288 is formed on a semiconductor substrate 286, and an SOI layer is formed on the buried oxide film 288. In the SOI layer, the diffusion regions 212 and 213 are formed and the other region is a body region 287.

[0160] By the sidewall memory cell as well, action and effect similar to those of the sidewall memory cell of the second embodiment are obtained. Further, junction capacitance between the diffusion regions 212 and 213 and the body region 287 can be remarkably reduced, so that higher-speed operation and lower power consumption of the device can be achieved.

[0161] Sixth Embodiment

[0162] A sidewall memory cell in a semiconductor memory device in a sixth embodiment has, as shown in FIG. 16, a configuration substantially similar to that of the sidewall memory cell of the second embodiment except that a P-type high-concentration region 291 is added adjacent to the channel sides of the N-type diffusion regions 212 and 213.

[0163] Specifically, the concentration of a P-type impurity (for example, boron) in the P-type high-concentration region 291 is higher than that of a P-type impurity in a region 292. Suitable P-type impurity concentration in the P-type high-concentration region 291 is, for example, about 5×1017 to 1×1019 cm−3. The P-type impurity concentration of the region 292 can be set to, for example, 5×1016 to 1×1018 cm−3.

[0164] By providing the P-type high-concentration region 291, the junction between the diffusion regions 212 and 213 and the semiconductor substrate 211 becomes sharp below the memory functional units 261 and 262. Consequently, hot carriers are easily generated in the writing and erasing operations, the voltage of the writing and erasing operations can be decreased or the writing operation and the erasing operation can be performed at high speed. Moreover, since the impurity concentration in the region 292 is relatively low, the threshold when the memory is in the erasing state is low, and the drain current is large. Consequently, the reading speed is improved. Therefore, the sidewall memory cell with low rewriting voltage or high rewriting speed and high reading speed can be obtained.

[0165] In FIG. 16, by providing the P-type high-concentration region 291 in the vicinity of the source/drain regions and below the memory functional unit (that is, not immediately below the gate electrode), the threshold of the whole transistor remarkably increases. The degree of increase is much higher than that in the case where the P-type high-concentration region 291 is positioned immediately below the gate electrode. In the case where write charges (electrons when the transistor is of the N-channel type) are accumulated in the memory functional unit, the difference becomes larger. On the other hand, in the case where sufficient erasing charges (positive holes when the transistor is of the N-channel type) are accumulated in the memory functional unit, the threshold of the whole transistor decreases to a threshold determined by the impurity concentration in the channel region (region 292) below the gate electrode. That is, the threshold in the erasing operation does not depend on the impurity concentration of the P-type high-concentration region 291 whereas the threshold in the writing operation is largely influenced. Therefore, by disposing the P-type high-concentration region 291 under the memory functional unit and in the vicinity of the source/drain regions, only the threshold in the writing operation largely fluctuates, and the memory effect (the difference between the threshold in the writing operation and that in the erasing operation) can be remarkably increased.

[0166] Seventh Embodiment

[0167] A sidewall memory cell in a semiconductor memory device of a seventh embodiment has a configuration substantially similar to that of the second embodiment except that, as shown in FIG. 17, the thickness (T1) of an insulating film separating the charge retaining film (silicon nitride film 242) and the channel region or well region is smaller than the thickness (T2) of the gate insulating film.

[0168] The thickness T2 of the gate insulating film 214 has the lower limit value from the demand of withstand voltage at the time of rewriting operation of the memory. However, the thickness T1 of the insulating film can be made smaller than T2 irrespective of the demand of withstand voltage.

[0169] The flexibility of designing with respect to T1 is high in the sidewall memory cell for the following reason.

[0170] In the sidewall memory cell, the insulating film for separating the charge retaining film and the channel region or well region is not sandwiched by the gate electrode and the channel region or well region. Consequently, to the insulating film for separating the charge retaining film and the channel region or well region, a high electric field acting between the gate electrode and the channel region or well region does not directly act, but a relatively low electric field spreading from the gate electrode in the lateral direction acts. Consequently, irrespective of the demand of withstand voltage to the gate insulating film, T1 can be made smaller than T2.

[0171] By making T1 thinner, injection of charges into the memory functional unit becomes easier, the voltage of the writing operation and the erasing operation is decreased or the writing operation and erasing operation can be performed at high speed. Since the amount of charges induced in the channel region or well region when charges are accumulated in the silicon nitride film 242 increases, the memory effect can be increased.

[0172] The electric lines of force in the memory functional unit include a short one which does not pass through the silicon nitride film 242 as shown by an arrow 284 in FIG. 13. On the relatively short electric line of force, electric field intensity is relatively high, so that the electric field along the electric line of power plays a big role in the rewriting operation. By reducing T1, the silicon nitride film 242 is positioned downward in the figure, and the electric line of force indicated by the arrow 283 passes through the silicon nitride film. Consequently, the effective dielectric constant in the memory functional unit along the electric line 284 of force increases, and the potential difference at both ends of the electric line of force can be further decreased. Therefore, a large part of the voltage applied to the gate electrode 217 is used to increase the electric field in the offset region, and the writing operation and the erasing operation become faster.

[0173] In contrast, for example, in an EEPROM typified by a flash memory, the insulating film separating the floating gate and the channel region or well region is sandwiched by the gate electrode (control gate) and the channel region or well region, so that a high electric field from the gate electrode directly acts. Therefore, in an EEPROM, the thickness of the insulating film separating the floating gate and the channel region or well region is regulated, and optimization of the function of the sidewall memory cell is inhibited.

[0174] As obvious from the above, by setting T1<T2, without deteriorating the withstand voltage performance of the memory, the voltage of the writing and erasing operations is decreased, or the writing operation and erasing operation are performed at high speed and, further, the memory effect can be increased.

[0175] More preferably, the thickness T1 of the insulating film is 0.8 nm or more at which uniformity or quality by a manufacturing process can be maintained at a predetermined level and which is the limitation that the retention characteristic does not deteriorate extremely.

[0176] Concretely, in the case of a liquid crystal driver LSI requiring high withstand voltage in a design rule, to drive the liquid crystal panel TFT, voltage of 15 to 18 V at the maximum is required, so that the gate oxide film cannot be thinned normally. In the case of mounting the nonvolatile memory of the present invention for image adjustment on the liquid crystal driver LSI, in the sidewall memory cell, the thickness of the insulating film separating the charge retaining film (silicon nitride film 242) and the channel region or well region can be designed optimally independently of the thickness of the gate insulating film. For example, the thickness can be individually set as T1=20 nm and T2=10 nm for a sidewall memory cell having a gate electrode length (word line width) of 250 nm, so that a sidewall memory cell having high writing efficiency can be realized (the reason why the short channel effect is not produced when T1 is larger than the thickness of a normal logic transistor is because the source and drain regions are offset from the gate electrode).

[0177] Eighth Embodiment

[0178] A sidewall memory cell in a semiconductor memory device of an eighth embodiment has a configuration substantially similar to that of the second embodiment except that, as shown in FIG. 18, the thickness (T1) of the insulating film separating the charge retaining film (silicon nitride film 242) and the channel region or well region is larger than the thickness (T2) of the gate insulating film.

[0179] The thickness T2 of the gate insulating film 214 has an upper limit value due to demand of preventing a short channel effect of the cell. However, the thickness T1 of the insulating film can be made larger than T2 irrespective of the demand of preventing the short channel effect. Specifically, when reduction in scaling progresses (when reduction in thickness of the gate insulating film progresses), the thickness of the insulating film separating the charge retaining film (silicon nitride film 242) and the channel region or well region can be designed optimally independent of the gate insulating film thickness. Thus, an effect that the memory functional unit does not disturb scaling is obtained.

[0180] The reason why flexibility of designing T1 is high in the sidewall memory cell is that, as described already, the insulating film separating the charge retaining film and the channel region or well region is not sandwiched by the gate electrode and the channel region or well region. Consequently, irrespective of the demand of preventing the short channel effect for the gate insulating film, T1 can be made thicker than T2.

[0181] By making T1 thicker, dissipation of charges accumulated in the memory functional unit can be prevented and the retention characteristic of the memory can be improved.

[0182] Therefore, by setting T1>T2, the retention characteristic can be improved without deteriorating the short channel effect of the memory.

[0183] The thickness T1 of the insulating film is, preferably, 20 nm or less in consideration of decrease in rewriting speed.

[0184] Concretely, in a conventional nonvolatile memory typified by a flash memory, a selection gate electrode serves as a write erase gate electrode, and a gate insulating film (including a floating gate) corresponding to the write erase gate electrode also serves as a charge accumulating film. Since a demand for size reduction (thinning of a film is indispensable to suppress short channel effect) and a demand for assuring reliability (to suppress leak of retained charges, the thickness of the insulating film separating the floating gate and the channel region or well region cannot be reduced to about 7 nm or less) are contradictory, it is difficult to reduce the size. Actually, according to the ITRS (International Technology Roadmap for Semiconductors), there is no prospect of reduction in a physical gate length of about 0.2 micron or less. In the sidewall memory cell, since T1 and T2 can be individually designed as described above, size reduction is made possible.

[0185] For example, for a sidewall memory cell having a gate electrode length (word line width) of 45 nm, T2=4 nm and T1=7 nm are individually set, and a sidewall memory cell in which the short channel effect is not produced can be realized. The reason why the short channel effect is not produced even when T2 is set to be thicker than the thickness of a normal logic transistor is because the source/drain regions are offset from the gate electrode.

[0186] Since the source/drain regions are offset from the gate electrode in the sidewall memory cell, as compared with a normal logic transistor, reduction in size is further facilitated.

[0187] Since the electrode for assisting writing and erasing does not exist in the upper part of the memory functional unit, a high electric field acting between the electrode for assisting writing and erasing and the channel region or well region does not directly act on the insulating film separating the charge retaining film and the channel region or well region, but only a relatively low electric field which spreads in the horizontal direction from the gate electrode acts. Consequently, the sidewall memory cell having a gate length which is reduced to be equal to or less than the gate length of a logic transistor of the same process generation can be realized.

[0188] Ninth Embodiment

[0189] A ninth embodiment relates to a change in the electric characteristic at the time of rewriting a sidewall memory cell of a semiconductor memory device.

[0190] In an N-channel type sidewall memory cell, when an amount of charges in a memory functional unit changes, a drain current (Id)-gate voltage (Vg) characteristic (actual measurement value) as shown in FIG. 19 is exhibited.

[0191] As obvious from FIG. 19, in the case of performing a writing operation in an erasing state (solid line), not only the threshold simply increases, but also the gradient of a graph remarkably decreases in a sub-threshold region.

[0192] Consequently, also in a region where a gate voltage (Vg) is relatively high, the drain current ratio between the erasing state and the writing state is high. For example, also at Vg=2.5V, the current ratio of two digits or more is maintained. The characteristic is largely different from that in the case of a flash memory (FIG. 22).

[0193] Appearance of such a characteristic is a peculiar phenomenon which occurs since the gate electrode and the diffusion region are offset from each other, and the gate electric field does not easily reach the offset region. When the sidewall memory cell is in a writing state, even when a positive voltage is applied to the gate electrode, an inversion layer is extremely hard to be formed in the offset region under the memory functional unit. This is the cause that the gradient of the Id-Vg curve is gentle in the sub-threshold region in the writing state.

[0194] On the other hand, when the sidewall memory cell is in an erasing state, electrons of high density are induced in the offset region. Further, when 0 V is applied to the gate electrode (that is, when the gate electrode is in an off state), electrons are not induced in the channel below the gate electrode (consequently, an off-state current is small). This is the cause that the gradient of the Id-Vg curve is sharp in the sub-threshold region in the erasing state, and current increasing rate (conductance) is high in the region of the threshold or more.

[0195] As obviously understood from the above, in the sidewall memory cell in the semiconductor memory device of the present invention, the drain current ratio between the writing operation and the erasing operation can be particularly made high.

[0196] Tenth Embodiment

[0197] A tenth embodiment relates to a semiconductor memory device in which a plurality of the sidewall memory cells of any of the first to ninth embodiments are arranged and, further, to a method of setting a reference cell necessary to read the semiconductor memory device.

[0198] FIG. 24 shows a part of a semiconductor memory device having a typical sidewall memory. As shown in FIG. 24, a memory array 20 has a plurality of sidewall memory cells 522 arranged in rows and columns. Although not specifically shown, a circuit for selecting a specific sidewall memory cell 522 by selecting a specific row and a specific column is provided.

[0199] The sidewall memory cell according to the present invention has a memory functional unit. By changing the amount of charges retained in the memory functional unit, the memory state is programmed to a writing state or an erasing state. The stored memory state can be detected by checking voltage (hereinafter, referred to as read voltage) generated when the drain current flowing in the sidewall memory cell flows into a load cell connected in series with the sidewall memory cell by a reading operation.

[0200] FIG. 25 shows distribution of threshold voltage in the writing state and the erasing state of the sidewall memory cell 522. The drain current flowing in the sidewall memory cell 522 is small in the state where charges are retained in the memory functional unit (writing state) and is large in the state where no charges are retained in the memory functional unit (erasing state). In order to read the memory state of the sidewall memory cell, a reference cell 529 is used. The reference cell 529 is constructed by a nonvolatile memory transistor according to the present invention and is the same cell as the sidewall memory cell 522. It is programmed so that an accurate charge amount corresponding to the reference state is accumulated in the memory functional unit in the reference cell 529. The memory state stored in the reference cell is detected by checking voltage (hereinafter, referred to as reference voltage) generated when the drain current flowing in the reference cell flows into a transistor, that is, a second load cell 532 connected in series with the reference cell by a reading operation. The memory state stored in the sidewall memory cell 522 is detected by checking voltage which is generated when the drain current flowing in the sidewall memory cell selected by the reading operation flows into a transistor, that is, a first load cell 527 connected in series with the selected sidewall memory cell 522. Specifically, by comparing the read voltage of the selected sidewall memory cell 522 with the reference voltage of the reference cell 529, the state of the selected sidewall memory cell 522 can be read. The sidewall memory cell 522 can be used as a cell for retaining two memory states of the writing state and the erasing state. In a memory array obtained by arranging therein the sidewall memory cells 522, the reference cell 529 is programmed to an intermediate state between the writing state and the erasing state of the sidewall memory cell 522. In the case of accessing the sidewall memory cell 522, it is constructed so that the reference cell 529 generates current of a specific value. As the microfabrication of the cell, there is a tendency that the read voltage decreases. In this case, two kinds of states of the memory become close to each other and a margin is reduced. Therefore, it is necessary to program the reference cell 529 accurately. Usually, voltage is applied to the gate terminal of the reference cell 529 and the gate voltage of the reference cell 529 is gradually increased until desired reference voltage is obtained on the basis of the drain current flowing in the reference cell 529. In such a manner, each reference cell 529 is programmed.

[0201] FIG. 24 shows typical arrangement of the sidewall memory cells 522 for reading a memory state. As understood from FIG. 24, the gate terminal of each sidewall memory cell 522 is connected to a word line 523, the source terminal is connected to the ground, and the drain terminal is connected to a bit line 524. To select a specific column, a transistor 525 is disposed for each column (hereinafter, referred to as a column selection transistor). An N-type field effect transistor 526 cascaded to the drain terminal of the column selection transistor 525 is disposed, and a drain bias is supplied to the sidewall memory cell 522. The source terminal of the N-type field effect transistor 526 is connected to the first load cell 527 connected to Vcc. As the first load cell, an N-type field effect transistor, a resistor element or the like is used.

[0202] In the case where voltage is applied to the gate of the sidewall memory cell 522 of a selected column, current which flows varies in accordance with the writing or erasing state of the sidewall memory cell 522. Current of the sidewall memory cell 522 in the writing state of high Vt is smaller than that of the sidewall memory cell 522 in the erasing state of low Vt. On the other hand, Vt is set so that intermediate current between the read current in the writing state and the read current in the erasing state flows into the reference cell 529.

[0203] When current flows in the sidewall memory cell 522, a voltage drop occurs between terminals of the first load cell 527. Voltage at the connection point between the first load cell 527 and the N-type field effect transistor 526 is applied to an input terminal SIN as one of input terminals of a sense amplifier 528. The other input terminal RIN of the sense amplifier 528 is connected to a similar current path in which cells 530, 531 and 532 and the reference cell 529 similar to the column selection transistor 525, N-type field effect transistor 526 and first load cell 527 are disposed. Hereinafter, the cell 532 will be referred to as a second load cell. By the drain current flowing in the reference cell 529, a voltage drop occurs between terminals of the second load cell 532. Voltage at the connection point between the second load cell 532 and the N-type field effect transistor 531 is applied to the other input terminal RIN of the sense amplifier 528. In the case where current flowing in the sidewall memory cell 522 by the reading operation is larger than the current flowing in the reference cell 529 (that is, in the case of the erasing state), the sense amplifier 528 generates a first output signal. In the case where the current flowing in the sidewall memory cell 522 is smaller than the current flowing in the reference cell 529 by the reading operation (that is, in the case of the writing state), the sense amplifier 528 generates a second output signal different from the first output signal. As described above, the current flowing in the reference cell 529 is sufficiently between the erasing state and the writing state but is, desirably, at the midpoint between the current flowing in the sidewall memory cell 522 in the writing state and the current flowing in the sidewall memory cell 522 in the erasing state for the following reason. When Vt of the reference cell 529 is set to a rather high value, although the current difference between Vt and the current flowing in the sidewall memory cell 522 in the erasing state of low Vt is large, the current difference between Vt and the current flowing in the sidewall memory cell 522 in the writing state of high Vt becomes small. On the other hand, when Vt of the reference cell 529 is set to a rather low value, although the current difference between Vt and the current flowing in the sidewall memory cell 522 in the writing state is large, the current difference between Vt and the current flowing in the sidewall memory cell 522 in the erasing state becomes small. In those cases, a design margin is small. Consequently, the reference cell 529 has to be programmed so as to retain charges of an accurate value. On the other hand, the sidewall memory cell 522 is programmed by a method quite different from the programming method of the reference cell 529. The sidewall memory cell 522 is programmed by applying voltage to each of the terminals of the sidewall memory cell 522 so that the read voltage of the sidewall memory cell 522, that is, the voltage of the input terminal SIN of the sense amplifier 528 is sufficiently low but is sufficient for the sense amplifier 528 to generate a second output signal during reading.

[0204] A method of verifying the memory state of the sidewall memory cell 522 and that of verifying the memory state of the reference cell 529 are quite different from each other. The memory state of the sidewall memory cell 522 is verified by normal reading operation using the reference cell 529 in the semiconductor memory cell and the sense amplifier 528. On the other hand, the reference cell 529 is programmed by using an external verification circuit in a process of manufacturing the semiconductor memory device.

[0205] If the method of programming the reference cell 529 is not devised, it can happen such that the memory state is not accurately reflected in an output of the sense amplifier 528 due to influences of a process, a temperature change and the like. Precision of measurement by the sense amplifier 528 is influenced by factors such as processes, temperatures and the like of the first load cell 527, second cell load cell 53, and related circuits. If the reference cell 529 is programmed without considering the factors, reading of the sidewall memory cell 522 cannot be sufficiently accurately performed. Therefore, in the case of programming the reference cell 529, it is desirable to check the memory state by using the sense amplifier 528 so that internal circuit elements exerting an influence on an output of the sense amplifier 528, that is, internal elements of the circuits exerting an influence on a voltage of the RIN terminal are considered.

[0206] A method of programming the reference cell 529 according to the present invention to a reference value will be described below.

[0207] In order to program the reference cell 529 very accurately by using the sense amplifier 528, accurate voltage is applied to the other input terminal SIN so as to be compared with the input terminal RIN to which the reference voltage is applied. It is also possible to directly connect an external test circuit system to a pad 535 or the like and apply voltage as shown in FIG. 24. In an embodiment of the circuit system according too the present invention, the digital/analog converter (not shown) may be provided in a part of a circuit in an integrated circuit including a memory array. It may be used to accurately generate voltage to be applied to the SIN terminal.

[0208] Further, means such as a switch 536 (preferably, a transistor switch) is provided for disconnecting the memory array from the sense amplifier 528 during programming of the reference cell 529. It can prevent the memory array from exerting an influence on the programming of the reference cell 529. Also by some methods other than the above-described method, the memory array can be disabled or disconnected from the sense amplifier 528 during programming of the reference cell.

[0209] To guarantee an accurate reference state for each reference cell 529, that is, to obtain accurate voltage to be applied to the SIN terminal at the time of programming the reference cell 529, the current-voltage characteristic of the first load cell 527 in a reading operation region of the sidewall memory cell 522 is carefully measured. As shown in FIG. 28, voltage corresponding to specific current in the operation region is determined. FIG. 28 shows a current-voltage characteristic in the case where the first load cell 527 is constructed by a resistor element. Different from general voltage applied during programming of the sidewall memory cell 522, those voltages can be measured very accurately. Therefore, voltage to be applied to the SIN terminal is measured accurately and the voltage can be applied to the SIN terminal of the sense amplifier 528 from the pad 535, a digital/analog converter (not shown), or a similar controlled voltage source. In such a manner, each reference cell 529 can be programmed so as to accumulate an accurate amount of charges.

[0210] Therefore, in the case of making programming so that some reference cells 529 supply desired current value I1, voltage V1 corresponding to the current value I1, which is determined from the current-voltage characteristic of FIG. 28, is applied to the SIN terminal via the switch 537. Charges are supplied to the terminal of the reference cell 529 until an output of the sense amplifier 528 is switched. Thus, the method of accumulating an accurate charge amount in the reference cell 529 is realized in consideration of the whole circuit 40 including, particularly, the circuit elements 530, 531 and 532. Therefore, when a specific value of current flowing in the sidewall memory cell 522 is known, a voltage drop which occurs in the first load cell 527 can be measured accurately, and the voltage can be applied to the SIN terminal and measured in comparison with the value of the voltage at the RIN terminal generated by the current flowing in the reference cell 529.

[0211] Thus, the method of programming the reference cell 529 in consideration of characteristics of cells related to the reading operation is provided. As described above, in the method of programming the reference cell 529 in the memory array to a reference state, the first load cell 527 in which read voltage of the sidewall memory cell 522 is generated is disconnected from the input to the sense amplifier 528, and an accurate voltage level which coincides with the voltage of the first load cell 527 in the selection state of the sidewall memory cell 522 is supplied to the sense amplifier 528 in place of the disconnected input. The pulse voltage to be applied to the gate terminal of the reference cell 529 is increased until the voltage from the reference cell 529 exceeds the voltage which coincides with the voltage of the load cell and an output of the sense amplifier 528 is switched.

[0212] As described above, a reference cell has to be programmed so that an accurate charge amount is accumulated in the reference cell 529. However, on the other hand, to reduce the manufacturing cost of the semiconductor memory device, a reference cell has to be programmed in short time. A method of programming a reference cell of a semiconductor memory device in short time in accordance with a process described below and testing the reference cell will be described. In the method of programming a reference cell, a controller such as a microprocessor provided in a part of the semiconductor memory device may be also used to operate a voltage switch or the like in the semiconductor device.

[0213] To program a reference cell in short time, a reference cell is programmed in two stages. In a first stage, a reference cell is adjusted so as to accumulate charges of a second charge amount smaller than a target charge amount at relatively high speed. First, a pulse of long time at high voltage is applied to the gate of the reference cell. Subsequently, a train of pulses of short time and of which voltages increase gradually is applied to the gate so that charges are accumulated gradually to the second charge amount as a target of the first stage. In the following second stage, a train of pulses of short time widths is applied at constant voltage to make the reference cell gradually shift to the target, and the reference cell is programmed until it exceeds the reference state as a final target. This method has the following advantages.

[0214] First, by the pulse applied to the gate first, the reference cell is shifted in a stroke to a saturation operation region exceeding a linear operation region. The influence of a change in the gate voltage in the saturation operation region is much smaller than that in the linear operation region. Therefore, accumulated charges can be stably, accurately controlled. Further, in the operation region, a change in the gate voltage and a change in the threshold voltage correspond to each other in a one-to-one manner, so that an increased amount of the accumulated charge amount by a change in the pulse voltage can be predicted very accurately.

[0215] Second, by repeatedly accumulating charges from a state of a small accumulated charge amount, the accumulated charges can be tested in each of processes. In the processes of the latter half of the programming, charges are accumulated more finely than precision required for the reference state. Consequently, a charge amount which is very close to the reference state is accumulated in a point in which an output of a sense amplifier is inverted in a testing process, so that accurate programming can be made in very short time.

[0216] It is particularly effective in the case where the voltage and time width of a pulse to be applied to the reference cell can be controlled by the controller in the semiconductor memory device.

[0217] FIG. 29 is a flowchart showing a method of programming a reference cell in the embodiment. It is assumed that a predetermined drain voltage Vd is applied in all of processes. In an embodiment, the voltage is 5.0 V.

[0218] In a first process, a first pulse for making a reference cell shift to the saturation operation region and accumulating charges close to the second charge amount into the reference cell is applied to the gate with Vg as the voltage (Vp) and tp1 as time width (tw) (step S01).

[0219] In an embodiment, the value of threshold voltage corresponding to the second charge amount is 150 mV which is lower than a final voltage Vt2 corresponding to the target charge amount. The reference voltage varies according to a target accumulated charge amount (or a corresponding threshold voltage). For example, in the case of accumulating charges corresponding to the threshold voltage of 1.5 V as an embodiment, it is sufficient to apply 5.0 V to the gate. Generally, the first pulse is the pulse having the widest time width among pulses applied to the gate of the reference cell. In the embodiment, the length of the pulse may be, for example, about 40 microseconds.

[0220] After the first pulse, the process shifts to a test for determining whether the amount of accumulated charges is smaller than the second charge amount or not. In the process, first, the first reference voltage Vt1 (150 mV in the embodiment) is applied to the SIN terminal of the sense amplifier via the switch 537 (step S02).

[0221] It is determined that whether or not an output of the sense amplifier is inverted from the second output state indicating that the charges accumulated in the reference cell is smaller than the target charge amount (step S03).

[0222] In the case where the accumulated charge amount has not reached the second charge amount after application of the first pulse, an output of the sense amplifier is not inverted. In this case, the second pulse is applied successively (step S04).

[0223] The time width tp2 of each of the pulses of the second pulse train is shorter than the time width tp1 of the first pulse (in the embodiment, tp2 is about ¼ or about 10 microseconds). In the pulse train, the voltages of the pulses increase only by Vstp pulse every pulse so that Vp increases by about 100 mV each time so that charges are accumulated to a desired level (in the embodiment, Vstp is about 100 mV). As a result, even when the accumulated charge amount of the nonvolatile memory cell becomes equal to or larger than the second charge amount, the voltage of the RIN terminal is smaller than the voltage Vt2 corresponding to the target charge amount by at least 50 mV.

[0224] After application of each of the pulses, the process (step S03) of determining whether the accumulated charge amount has reached the second charge amount or not on the basis of inversion of an output of the sense amplifier is repeated.

[0225] When an output of the sense amplifier has inverted, that is, when the accumulated charge amount reaches the second charge amount, the routine shifts to a process of testing whether the accumulated charge amount has reached the target charge amount or not. The process is performed by applying voltage corresponding to the target charge amount to the SIN terminal and determining whether the output of the sense amplifier is inverted or not.

[0226] In the process, first, the second reference voltage Vt2 is applied via the switch 537 to the SIN terminal of the sense amplifier (step S05).

[0227] It is determined that whether an output of the sense amplifier is inverted from the second output state or not (step S06).

[0228] In the case where the accumulated charge amount has not reached the target charge amount, an output of the sense amplifier is not inverted from the second output state. In this case, subsequently, the third pulse train is applied to the gate (step S07).

[0229] The time width tp3 of each of pulses of the third pulse train is time width of about {fraction (1/10)} of the second pulse train (in the embodiment, 1 microsecond), and the pulse voltage is equal to the voltage of the pulse of the second pulse train which is applied at last. Generally, the third pulse train is selected so as to shift the voltage at the RIN terminal in steps each of which is about 10 mV or less to the voltage Vt2 corresponding to the target charge amount. By repeatedly applying the third pulse, the accumulated charges in the reference cell are becoming closer to the target charge amount. Therefore, when an output of the sense amplifier is inverted and it shows that the accumulated charge amount is equal to or larger than the target charge amount, the RIN terminal voltage lies within 10 mV of the final value. The third charge amount as the accumulated charge amount at this time is very close to the target charge amount.

[0230] After reaching the target, the program is finished. In such a manner, the reference cell can reach its reference charge amount very promptly, and the achieved value is very accurate. In the embodiment, each reference cell reaches the final value in about 1 ms. Numerical values shown as an embodiment are just an example, and the present invention is not limited to the numerical values.

[0231] In the embodiment, the case of storing two values of the state where charges are retained in the memory functional unit and the state where no charges are retained, that is, the writing state and the erasing state has been described. By controlling the charge amount accumulated in the memory functional unit, information of four or larger values can be also stored. For example, by setting the distribution of threshold voltages in the memory states of four values of the sidewall memory cell 522 as shown in FIG. 27, four values can be stored. In this case, three kinds of states, specifically, a state between the states 0 and 1, a state between the states 1 and 2, and a state between the states 2 and 3 are necessary for the reference voltage. FIG. 26 shows a circuit block for reading when information of four values is stored in the sidewall memory cell. It is sufficient to arrange the reference cells 529 and selection transistors 530 of the reference cells, each of the necessary reference number. In the case of four values, it is sufficient to arrange three reference cells 529 and three selection transistors 530. The reference cell 529 can be set in a manner similar to the method in the case of two values so that thresholds of three reference cell transistors become Vt between the states 0 and 1, Vt between the states 1 and 2, and Vt between the states 2 and 3.

[0232] Eleventh Embodiment

[0233] As an application example of the semiconductor memory device, for example, as shown in FIG. 20, a rewritable nonvolatile memory for image adjustment of a liquid crystal panel can be mentioned.

[0234] A liquid crystal panel 1001 is driven by a liquid crystal driver 1002. In the liquid crystal driver 1002, a nonvolatile memory 1003, an SRAM 1004 and a liquid crystal driver circuit 1005 are provided. The nonvolatile memory 1003 is constructed by the sidewall memory cell of the present invention, more preferably, any of the semiconductor memory devices of the tenth embodiments. The nonvolatile memory 1003 can be rewritten from the outside.

[0235] Information stored in the nonvolatile memory 1003 is transferred to the SRAM 1004 at the time of turn-on of the power source of an apparatus. The liquid crystal driver circuit 1005 can read stored information from the SRAM 1004 as necessary. By providing the SRAM, high reading speed of stored information can be achieved.

[0236] The liquid crystal driver 1002 may be externally attached to the liquid crystal panel 1001 as shown in FIG. 20 or formed on the liquid crystal panel 1001.

[0237] In a liquid crystal panel, tones displayed by applying voltages in multiple grades to pixels are changed. The relation between the given voltage and the displayed tone varies according to products. Consequently, information for correcting variations in each product after completion of the product is stored and correction is made on the basis of the information, thereby enabling the picture qualities of products to be made uniform. It is therefore preferable to mound a rewritable nonvolatile memory for storing correction information. As the nonvolatile memory, it is preferable to use the sidewall memory cell. Particularly, it is preferable to use any of the semiconductor memory devices of the tenth embodiments in which sidewall memory cells of the present invention are integrated.

[0238] Twelfth Embodiment

[0239] FIG. 21 shows a cellular telephone as a portable electronic apparatus incorporating the above-described semiconductor memory device.

[0240] The cellular telephone is mainly constructed by a control circuit 811, a battery 812, an RF (Radio Frequency) circuit 813, a display 814, an antenna 815, a signal line 816 and a power source line 817. The control circuit 811 incorporates therein the above-described semiconductor memory device according to the present invention. Herein, the control circuit 811 is preferably an integrated circuit, in which devices of the same configuration serve as a memory circuit device and a logic circuit device, as described in the tenth embodiment. Consequently, the integrated circuit can be easily fabricated, and thus, the fabrication cost of the portable electronic apparatus can be remarkably reduced.

[0241] In this manner, it is possible to increase the operation speed of the portable electronic apparatus and reduce the fabrication cost, so as to provide the inexpensive portable electronic apparatus having high reliability and high performance by using, for the portable electronic apparatus, the semiconductor memory device which facilitates the combination process between a memory and a logic circuit and achieves a reading operation at a high speed.

[0242] According to the present invention, at the time of programming a nonvolatile memory cell, charges of a first charge amount which is smaller than a target charge amount are accumulated, after that, a pulse train is applied to a second charge amount which is smaller than the target charge amount and is larger than the first charge amount and, further, a short pulse train is applied to a third charge amount which is within an allowable error range of the target charge amount. Consequently, a reference cell for storing a reference to be compared in order to read a memory state can be accurately programmed in short time to a state where the reference is stored.

[0243] When the first voltage and time width are set so as to make the nonvolatile memory cell reach a saturation operation region, the reference cell is shifted by the first pulse in a stroke to a saturation operation region exceeding a linear operation region, so that accumulated charges can be controlled stably and accurately. Further, in the operation region, a change in the gate voltage and a change in the threshold voltage correspond to each other in a one-to-one manner, so that an increase amount of the accumulated charge amount by a change in the pulse voltage can be predicted very accurately.

[0244] By providing the nonvolatile memory cell with a diffusion region which is disposed so as to be offset from the gate electrode, when voltage is applied to the gate electrode, easiness of inversion of the offset region below the charge retaining film largely changes in accordance with an amount of charges accumulated in the memory functional unit. Thus, the memory effect increases, and a short channel effect can be reduced.

[0245] In the sidewall memory cell as a component of the semiconductor memory device, a memory function of the memory functional unit and a transistor operation function of the gate insulating film are separated from each other. Consequently, it is easy to suppress the short channel effect by thinning the gate insulating film while making the sufficient memory function remained. Further, a value of current flowing between the diffusion regions changes by rewriting more largely than that in the case of an EEPROM. Therefore, it facilitates discrimination between the writing state and the erasing state of the semiconductor memory device.

[0246] Further, the sidewall memory cell can be formed by a process which has very high affinity with a normal transistor forming process on the basis of the configuration. Therefore, as compared with the case of using a conventional flash memory as a nonvolatile memory cell and forming the sidewall memory cell together with a peripheral circuit which is usually made by a transistor, the number of masks and the number of processes can be dramatically reduced.

[0247] Consequently, the yield in production of a chip on which the sidewall memory cell and the peripheral circuit are formed together can be improved. Because of this, the manufacturing cost is reduced and, accordingly, a very-reliable, cheap semiconductor memory device can be obtained.

[0248] When the semiconductor memory device of the present invention may have a sidewall memory cell including a film having a surface almost parallel to the surface of the gate insulating film and having the function of retaining charges, variations in the memory effect can be suppressed. Further, a large design margin for variations in the sidewall memory cell can be set, so that designing is facilitated.

[0249] When the semiconductor memory device of the present invention may have the sidewall memory cell in which the memory functional unit has a charge retaining film which is an insulating film, a charge leakage is prevented, and an excellent retention characteristic is obtained.

[0250] When the semiconductor memory device of the present invention includes a sidewall memory cell including an insulating film for separating a film having a surface almost parallel to a surface of a gate insulating film and having the function of retaining charges from the channel region or the semiconductor layer, and the thickness of the insulating film is larger than that of the gate insulating film and is 20 nm or less, an excellent retention characteristic is obtained.

[0251] When a programming method of a semiconductor memory device of the present invention includes the steps of: generating comparative voltage substantially equal to voltage obtained from the current-voltage characteristic of the first load cell in accordance with the reference current; and compensating variations in the first load cell by programming the reference cell so that the reference voltage of the second load cell becomes equal to the comparative voltage, the reference cell can be programmed accurately to a state where a reference is stored.

Claims

1. A method for programming a nonvolatile memory cell comprising the steps of: applying a first pulse having a first time width at a first voltage to a nonvolatile memory cell to accumulate a first amount of charge which is smaller than a target amount of charge; testing the nonvolatile memory cell to determine whether or not the amount of charge accumulated in the memory cell is larger than a second amount of charge; in the case where the accumulated first amount of charge is smaller than the second amount of charge, applying a second pulse train to the nonvolatile memory cell, the second pulse train having pulses each having a voltage by which the second amount of charge can be accumulated with a second time width narrower than the first time width so that charges in an amount close to the second amount of charge are accumulated in the nonvolatile memory cell; after applying each pulse of the second pulse train, testing the nonvolatile memory cell to determine whether or not the memory cell retains charges larger than the second amount of charge; when the nonvolatile memory cell is determined as retaining charges larger than the second amount of charge; applying a third pulse train having pulses each having a third voltage with a third time width narrower than the second time width to the nonvolatile memory cell until charges within an allowable error range of the target amount of charge are stored in the memory cell; and wherein the nonvolatile memory cell includes a gate electrode formed on a semiconductor layer via a gate insulating film, a channel region disposed under the gate electrode, a source and a drain as diffusion regions disposed on both sides of the channel region and having a conductive type opposite to that of the channel region, and memory functional units formed on both sides of the gate electrode and having a function of retaining charges.

2. The method according to claim 1, wherein all of the pulses are applied to the gate electrode of the nonvolatile memory cell while the drain and the source are maintained in a predetermined state.

3. The method according to claim 1, wherein the first pulse has a first voltage sufficient to make the nonvolatile memory cell to be in a saturation region, each pulse of the second pulse train has a voltage larger than a voltage of the immediately preceding pulse, and a voltage of the third pulse train is equal to a voltage of a final pulse of the second pulse train.

4. A method for programming a nonvolatile memory cell comprising the steps of: applying a first pulse to a gate electrode while maintaining a drain and a source of the nonvolatile memory cell in a predetermined state, the first pulse having a first time width at a first voltage sufficient to make a nonvolatile memory cell be in a saturation region so that a first amount of charge which is smaller than a target amount of charge is accumulated; testing the nonvolatile memory cell to determine whether or not the amount of charge accumulated in the memory cell is larger than a second amount of charge; in the case where the nonvolatile memory cell retains charges smaller than the second amount of charge, applying a second pulse train to the gate electrode of the nonvolatile memory cell, the second pulse train having a voltage by which the second amount of charge can be accumulated with a second time width narrower than the first time width so that charges close to the second amount of charge are accumulated in the nonvolatile memory cell; after applying each of the pulses of the second pulse train, testing the nonvolatile memory cell to determine whether or not the memory cell retains charges larger than the second amount of charge; in the case where the nonvolatile memory cell is determined as retaining charges larger than the second amount of charge, applying a third pulse train to the nonvolatile memory cell, the third pulse train having a third voltage with a third time width narrower than the second time width of each pulse so that a third amount of charge applied by pulses of the third pulse train is substantially equal to an amount of charge in an allowable error range of the target amount of charge; and after applying each pulse of the third pulse train, testing the nonvolatile memory cell to determine whether or not the memory cell retains charges larger than the target amount of charge, wherein in the nonvolatile memory cell, the gate electrode is formed on a semiconductor layer via a gate insulating film and has a channel region disposed under the gate electrode, the source and the drain as diffusion regions are disposed on both sides of the channel region and have a conductive type opposite to that of the channel region, and the gate electrode has memory functional units having a function of retaining charges formed on both sides thereof.

5. The method according to claim 4, wherein each pulse of the second pulse train has a voltage larger than a voltage of the immediately preceding pulse, and the third voltage is equal to a voltage of a final pulse of the second pulse train.

6. A method for programming a nonvolatile memory cell comprising the steps of: applying a first pulse having a first time width to a nonvolatile memory cell so that the nonvolatile memory cell retains a first amount of charge which is smaller than a target amount of charge for accumulating charges close to the target amount of charge in the nonvolatile memory cell; applying a second pulse to the nonvolatile memory cell, the second pulse train having pulses each having a second amount of charge with a second time width which is narrower than the first time width so that the nonvolatile memory cell retains the second amount of charge which is smaller than the target amount of charge and larger than the first amount of charge; and applying a third pulse train to the nonvolatile memory cell so that the nonvolatile memory cell retains a third amount of charge within an allowable error range of the target amount of charge, wherein the nonvolatile memory cell includes a gate electrode formed on a semiconductor layer via a gate insulating film, a channel region disposed under the gate electrode, a source and a drain as diffusion regions disposed on both sides of the channel region and having a conductive type opposite to that of the channel region, and memory functional units formed on both sides of the gate electrode and having a function of retaining charges.

7. The method according to claim 6, further comprising the steps of: after applying the first pulse, testing the nonvolatile memory cell to determine whether or not the first amount of charge exceeds the second amount of charge; in the case where the first amount of charge exceeds the second amount of charge, not executing a step of applying the second pulse train, and in the case where the first amount of charge is below the second amount of charge, after applying each pulse of the second pulse train, testing the nonvolatile memory cell to determine whether or not the memory cell retains the second amount of charge, and continuously applying the second pulse train until the nonvolatile memory cell is determined as retaining the second amount of charge; and after applying each pulse of the third pulse train, testing the nonvolatile memory cell to determine whether or not the memory cell retains the third amount of charge, and continuously applying the third pulse train until the nonvolatile memory cell is determined as retaining the third amount of charge.

8. The method according to claim 7, wherein a voltage of each pulse of the second pulse train is larger than a voltage of the immediately preceding pulse.

9. The method according to claim 6, wherein the first pulse has a first voltage, and the first voltage and first time width are selected so that the nonvolatile memory cell lies in a saturation region.

10. The method according to any one of claims 1 to 9, wherein the diffusion regions of the nonvolatile memory cell are disposed so as to be offset from the gate electrode.

11. The method according to any one of claims 1 to 9, wherein the nonvolatile memory cell includes a film having a lower surface which extends substantially parallel to a lower surface of the gate insulating film and having a function of retaining charges.

12. The method according to any one of claims 1 to 9, wherein the memory functional units of the nonvolatile memory cell each have a charge retaining film which is an insulating film.

13. The method according to any one of claims 1 to 9, wherein the nonvolatile memory cell includes a film having a lower surface which extends substantialy parallel to a lower surface of the gate insulating film and having a function of retaining charges, and an insulating film for separating the film from the channel region or the semiconductor layer, and a thickness of the insulating film is larger than that of the gate insulating film and is 20 nm or less.

14. A semiconductor memory device comprising a circuit for programming a nonvolatile memory cell, the programming circuit comprising: means for applying a first pulse having a first time width to a nonvolatile memory cell to accumulate a first amount of charge smaller than a target amount of charge in a nonvolatile memory cell so that the nonvolatile memory cell retains charges close to the target amount of charge; means for applying a second pulse train to the nonvolatile memory cell, the second pulse train having pulses each having a second time width which is narrower than the first time width so that the nonvolatile memory cell retains a second amount of charge which is smaller than the target amount of charge and larger than the first amount of charge; and means for applying a third pulse train to the nonvolatile memory device, the third pulse train having pulses each having a third time width which is narrower than the second time width so that the nonvolatile memory cell retains a third amount of charge within an allowable error range of the target amount of charge, wherein the nonvolatile memory cell includes a gate electrode formed on a semiconductor layer via a gate insulating film, a channel region disposed under the gate electrode, a source and a drain as diffusion regions disposed on both sides of the channel region and having a conductive type opposite to that of the channel region, and memory functional units formed on both sides of the gate electrode and having a function of retaining charges.

15. The semiconductor memory device according to claim 14, wherein the programming circuit further comprises: first means for testing a nonvolatile memory cell to determine, after applying the first pulse, whether or not the first amount of charge exceeds the second amount of charge; second means for testing the nonvolatile memory cell to determine, after applying pulses of the second pulse train, whether or not the nonvolatile memory cell retains the second amount of charge; and third means for testing the nonvolatile memory cell to determine, after applying pulses of the third pulse train, whether or not the nonvolatile memory cell retains the third amount of charge, and wherein, in the case where the first amount of charge exceeds the second amount of charge as a result of a test by the first means, the application of the second pulse train is not executed, in the case where the nonvolatile memory cell is determined as not retaining the second amount of charge as a result of a test by the second test means, the second pulse train is continuously applied until the nonvolatile memory cell is determined as retaining the second amount of charge, and in the case where the nonvolatile memory cell is determined as not retaining the third amount of charge as a result of a test by the third test means, the third pulse train is continuously applied until the nonvolatile memory cell is determined as retaining the third amount of charge.

16. The semiconductor memory device according to claim 15, wherein a voltage of each pulse of the second pulse train is larger than a voltage of the immediately preceding pulse.

17. The semiconductor memory device according to claim 14, wherein the first pulse has a first voltage, and the first voltage and first time width are selected so that the nonvolatile memory cell lies in a saturation region.

18. The semiconductor memory device according to any one of claims 14 to 17, wherein the diffusion regions of the nonvolatile memory cell are disposed so as to be offset from the gate electrode.

19. The semiconductor memory device according to any one of claims 14 to 17, wherein the nonvolatile memory cell includes a film having a lower surface which extends substantially parallel to a lower surface of the gate insulating film and having a function of retaining charges.

20. The semiconductor memory device according to any one of claims 14 to 17, wherein the memory functional units of the nonvolatile memory cell each have a charge retaining film which is an insulating film.

21. A portable electronic apparatus comprising the semiconductor memory device according to any one of claims 14 to 17.