JP2007208273A - Semiconductor device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a nonvolatile memory capable of being integrally formed with other semiconductor devices. <P>SOLUTION: Memory elements, switching elements, and other peripheral circuit for constituting a nonvolatile memory are integrally formed on a substrate using TFTs. Since the thickness of the semiconductor active layers of memory element TFTs are thinner than the thickness of the semiconductor active layers of other TFTs, impact ionization easily occurs in the channel region of the memory element TFTs. This makes it possible to implement the low voltage writing/erasing of the memory elements. There is provided a non volatile memory in which deterioration hardly occurs and which can be downsized. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a nonvolatile memory of a thin film transistor formed using SOI (Silicon On Insulator) technology. In particular, the present invention relates to an EEPROM (Electrically Erasable and Programmable Read Only Memory) integrally formed on an insulating substrate together with peripheral circuits such as a drive circuit. Further, the term “Silicon” here refers to a single crystal or a substantially single crystal.

  In recent years, with the miniaturization of semiconductor devices, high performance, high storage capacity, and small memory have been required. Currently, a semiconductor nonvolatile memory made of a magnetic disk or bulk silicon is most often used as a storage device of a semiconductor device.

  The magnetic disk is one of the largest storage capacities used in semiconductor devices, but has the disadvantages that it is difficult to reduce the size and the writing / reading speed is slow.

  On the other hand, the semiconductor nonvolatile memory is currently inferior to the magnetic disk in terms of storage capacity, but its writing / reading speed is several tens of times that of the magnetic disk. Semiconductor non-volatile memories have been developed that have sufficient performance with respect to the number of rewrites and data retention time. Therefore, recently, there has been an increase in the use of semiconductor memory as an alternative to magnetic disks.

  However, conventionally, a semiconductor nonvolatile memory is manufactured using bulk silicon and housed in a package. Therefore, when such a semiconductor nonvolatile memory is mounted on a semiconductor device, the number of processes is increased and the package size is increased. This has hindered miniaturization of semiconductor devices.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a non-volatile memory that can be integrally formed with components of other semiconductor devices and can be miniaturized.

According to an embodiment of the present invention,
A nonvolatile memory in which memory cells including memory TFTs and switching TFTs are arranged in a matrix,
The memory TFT includes at least a semiconductor active layer formed on an insulating substrate, a gate insulating film, a floating gate electrode, an anodized film obtained by anodizing the floating gate electrode, and a control gate electrode. The switching TFT includes at least a semiconductor active layer formed on the insulating substrate, a gate insulating film, and a gate electrode;
The nonvolatile memory is characterized in that the memory TFT and the switching TFT are integrally formed on the insulating substrate, and the thickness of the semiconductor active layer of the memory TFT is smaller than the thickness of the semiconductor active layer of the switching TFT. Is provided. This achieves the above object.

  The thickness of the semiconductor active layer of the memory TFT and the switching TFT may be less than 150 nm.

  The thickness of the semiconductor active layer of the memory TFT may be 1 to 50 nm, and the thickness of the semiconductor active layer of the switching TFT may be 40 to 100 nm.

  The thickness of the semiconductor active layer of the memory TFT may be 10 to 40 nm.

  The thickness of the semiconductor active layer of the memory TFT may be such that impact ionization is more likely to occur than the thickness of the semiconductor active layer of the switching TFT.

  The tunnel current flowing in the semiconductor active layer of the memory TFT may be more than twice the tunnel current flowing in the semiconductor active layer of the switching TFT.

Also, according to an embodiment of the present invention,
A nonvolatile memory in which memory cells including memory TFTs and switching TFTs are arranged in a matrix,
The memory TFT includes at least a control gate electrode formed on an insulating substrate, a first insulating film, a floating gate electrode, a second insulating film, and a semiconductor active layer,
The switching TFT includes at least a gate electrode formed on the insulating substrate, a first insulating film, and a semiconductor active layer,
The nonvolatile memory is characterized in that the memory TFT and the switching TFT are integrally formed on the insulating substrate, and the thickness of the semiconductor active layer of the memory TFT is smaller than the thickness of the semiconductor active layer of the switching TFT. Is provided. This achieves the above object.

  The thickness of the semiconductor active layer of the memory TFT and the switching TFT may be less than 150 nm.

  The thickness of the semiconductor active layer of the memory TFT may be 1 to 50 nm, and the thickness of the semiconductor active layer of the switching TFT may be 40 to 100 nm.

  The semiconductor active layer of the memory TFT may have a thickness of 10 to 40 nm, and the semiconductor active layer of the switching TFT may have a thickness of 40 to 100 nm.

  The thickness of the semiconductor active layer of the memory TFT may be such that impact ionization is more likely to occur than the thickness of the semiconductor active layer of the switching TFT.

  The tunnel current flowing in the semiconductor active layer of the memory TFT may be more than twice the tunnel current flowing in the semiconductor active layer of the switching TFT.

Also, according to an embodiment of the present invention,
Forming an amorphous silicon film having a first thickness and an amorphous silicon film having a second thickness on an insulating substrate;
The amorphous silicon film having the first thickness and the amorphous silicon film having the second thickness are crystallized to obtain a polycrystalline silicon film having the first thickness, Forming a polycrystalline silicon film having a thickness;
Forming a memory TFT on the first polycrystalline silicon film and forming a switching TFT on the second polycrystalline silicon film;
A method for manufacturing a non-volatile memory including:
There is provided a method for manufacturing a nonvolatile memory, wherein the first thickness is thinner than the second thickness. This achieves the above object.

  The thickness of the semiconductor active layer of the memory TFT and the switching TFT may be less than 150 nm.

  The first thickness may be 1 to 50 nm, and the second thickness may be 40 to 100 nm.

  The first thickness may be 10 to 40 nm.

  The thickness of the semiconductor active layer of the memory TFT may be such that impact ionization is more likely to occur than the thickness of the semiconductor active layer of the switching TFT.

  The tunnel current flowing in the semiconductor active layer of the memory TFT may be more than twice the tunnel current flowing in the semiconductor active layer of the switching TFT.

  According to the present invention, the non-volatile memory is integrally formed on the same substrate as the peripheral circuit such as the drive circuit, and the size can be reduced.

  Further, according to the present invention, since the semiconductor active layer of the nonvolatile memory is relatively thin, impact ionization is likely to occur, and a nonvolatile memory that is driven at a low voltage and has little deterioration is realized.

  Furthermore, since the nonvolatile memory of the present invention can be formed integrally with the components of the semiconductor device, the semiconductor device can be reduced in size.

  In this embodiment, a nonvolatile memory formed on an insulating substrate, particularly an EEPROM will be described. The EEPROM of this embodiment is integrally formed on an insulating substrate together with peripheral circuits such as a drive circuit.

  Please refer to FIG. FIG. 1 shows a circuit diagram of a 4k-bit EEPROM of this embodiment. The 4k-bit EEPROM of this embodiment is configured by a plurality of electrically erasable memory elements Tr1, a plurality of switching elements Tr2, X and Y address decoders 101 and 102, and other peripheral circuits 103 and 104. Other peripheral circuits include an address buffer circuit and a control logic circuit, and are provided as necessary. In FIG. 1, a memory element (storage element) in which each bit information is recorded is indicated by Tr1. Tr1 is a P-channel nonvolatile memory having a floating gate. Tr2 is an N-channel switching element Tr2.

  The drain electrodes of the two TFTs (Tr1 and Tr2) are connected in series with each other, and this serial connection circuit constitutes a 1-bit memory cell. In this embodiment, the memory cells are arranged in a matrix of 64 vertical × 64 horizontal. Since each memory cell can store 1-bit information, the EEPROM of this embodiment has a storage capacity of 4096 bits (= about 4 kbits). In this embodiment, an EEPROM having a storage capacity of 4096 bits will be described. However, the present invention can be applied to an EEPROM having any storage capacity.

  The memory cells arranged in each column are connected at both ends to signal lines labeled A0, B0 to A63, B63. In the memory cells arranged in each row, the gate electrodes of the memory cells are connected to the signal lines C0, D0 to C63 to D63. As shown in FIG. 1, in this embodiment, each memory cell constituting the 4k-bit EEPROM is given a reference numeral (0, 0), (1, 0), (63, 63).

  The signal lines A0, B0 to A63, B63, and C0, D0 to C63 to D63 are connected to the X address decoder 101 and the Y address decoder 102, respectively. A specific memory cell is specified by the X address decoder 101 and the Y address decoder 102, and data is written, read or erased.

  Next, the configuration of the memory cell of this embodiment will be described with reference to FIG. FIG. 2 shows a cross-sectional view of the memory cell of this embodiment. In FIG. 2, the left element is the memory element Tr1, and the right element is the switching element Tr2. The semiconductor active layer 202 of Tr1 includes source / drain regions 203 and 204 and a channel region 205. On the other hand, the semiconductor active layer 206 of Tr 2 includes source / drain regions 207 and 208, a low concentration impurity region 209, and a channel region 210. 211 and 212 are gate insulating films. Reference numeral 213 denotes a floating gate electrode. Reference numerals 214 and 218 denote anodic oxide films. Reference numeral 215 denotes a control gate electrode. Reference numerals 216, 220, and 219 denote source / drain electrodes. Reference numeral 221 denotes an interlayer insulating film.

  As shown in FIG. 2, the thickness d1 of the semiconductor active layer of the memory element Tr1 is different from the thickness d2 of the semiconductor active layer of the switching element Tr2, and d1 <d2. By doing so, impact ionization (impact ionization or impact ionization) in the semiconductor active layer of the memory element Tr1 is more likely to occur, and charge injection to the floating gate electrode of Tr1 is likely to occur. The tunnel current flowing in the semiconductor active layer of the memory element is preferably at least twice as large as the tunnel current flowing in the semiconductor active layer of the switching element. As a result, the memory element can be written / erased at a low voltage, so that the memory element is hardly deteriorated with respect to the number of times the memory element is written.

  Note that the thicknesses of the semiconductor active layers of the TFTs constituting the X and Y address decoders 101 and 102 and the TFTs constituting other peripheral circuits are the same as the thickness of the switching TFT Tr2.

  Here, the operation of the EEPROM of the present embodiment will be described by taking the memory cell (1, 1) in FIG. 1 as an example.

  First, when data is written to the memory element (1, 1), a voltage of −5 V is applied to A1. A voltage of 5V is also applied to the signal line D1. Therefore, when the signal line B1 is set to GND and a high voltage of about 20 V is applied to the signal line C1, carriers (holes in this case) moving in the channel region of Tr1 are accelerated, and weak avalanche collapse or impact ionization occurs. And a large amount of hot carriers (electrons) in a high energy state are generated. This carrier is injected into the gate insulating film and trapped by the floating gate electrode. In this way, charges are stored in the floating gate electrode of Tr1. As the charge is stored in the floating gate electrode, the threshold voltage of Tr1 changes.

  Next, when reading data from the memory element (1, 1), 0V is applied to the signal line C1, and 5V is applied to D1. When B1 is set to GND, the threshold voltage of Tr1 changes depending on whether charges are stored in the floating gate electrode or not, and the stored signal is read out from A1. Become.

  Next, when erasing data stored in the memory element (1, 1), 5 V is applied to the signal line D1, and the signal line B1 is set to GND. When a voltage of about −20 V is applied to the signal line C1, electrons trapped in the floating gate electrode are injected into the drain region. Therefore, the stored data is erased.

  The above operations are summarized in the table below.

  Note that the voltage applied to the memory element depends on the thickness of the semiconductor active layer of the memory element, the capacitance between the control gate electrode and the floating electrode, and the like. Therefore, the operating voltage of the memory element is not limited to the voltage described above.

  In the EEPROM, the number of times of rewriting and the information holding time are important. In order to increase the number of times of rewriting, it is required to reduce the voltage applied to the control electrode of the memory element. Since the thickness of the semiconductor active layer of the memory element of this embodiment is larger than the thickness of the semiconductor active layer of the TFT constituting the switching TFT or the address decoder, impact ionization is likely to occur, and the voltage applied to the control electrode Can be lowered.

  In this embodiment, when data is written / erased in the memory element, a voltage of 20 V is not applied to the control electrode of the memory element at a time, but a voltage lower than this voltage is applied to a plurality of pulses. It is possible to prevent the deterioration of the element by applying the voltage at.

  In addition, since the TFT constituting the EEPROM of this embodiment requires high characteristics in terms of mobility, threshold voltage, etc., a TFT having an amorphous silicon semiconductor active layer that is often used conventionally is not sufficient. Absent. Therefore, a method for manufacturing a TFT exhibiting high characteristics as described above will be described below. According to the following manufacturing method, a TFT with high characteristics can be manufactured, and the EEPROM of this embodiment can be realized.

  A method for manufacturing the EEPROM of this embodiment will be described with reference to FIGS. 3 to 7 show a typical CMOS circuit as a TFT constituting the EEPROM of this embodiment as a circuit constituting a memory element constituting a memory cell, a switching element, and an address decoder and other peripheral circuits. An explanation will be given by taking two TFTs constituting the above as an example.

  In addition, according to the nonvolatile memory manufacturing method described below, it is understood that any semiconductor device that can be manufactured using thin film technology and the nonvolatile memory of the present invention can be integrally formed.

  Please refer to FIG. First, a quartz substrate 301 is prepared as a substrate having an insulating surface. A silicon substrate on which a thermal oxide film is formed can be used instead of the quartz substrate. Alternatively, a method may be employed in which an amorphous silicon film is once formed on a quartz substrate and is completely thermally oxidized to form an insulating film. Further, a quartz substrate or a ceramic substrate on which a silicon nitride film is formed as an insulating film may be used.

Next, an amorphous silicon film 302 is formed to a thickness of 25 nm (FIG. 3A). In this embodiment, the film is formed by a low pressure thermal CVD method and formed according to the following conditions.
Deposition temperature: 465 ° C
Deposition pressure: 0.5 torr
Deposition gas: He (helium) 300 sccm
Si ₂ H ₆ (disilane) 250 sccm

Next, a resist film is formed and patterned to form a mask 304 (FIG. 3B). After that, the amorphous silicon film 303 is etched to form an amorphous silicon film 304 partially formed on the substrate (FIG. 3C). Note that the amorphous silicon film 303 may be etched by either dry etching or wet etching. In the case of dry etching, CF ₄ + O ₂ may be used, and in the case of wet etching, fluorine acid + nitric acid may be used.

  Next, an amorphous silicon film is again formed to a thickness of 50 nm by the method described above, and amorphous silicon films 305 and 306 as shown in FIG. 3D are formed. Here, the final film thickness (thickness considering the film reduction after thermal oxidation) was adjusted so that the amorphous silicon film 305 was 50 nm and the amorphous silicon film 306 was 75 nm.

  Note that it is desirable to clean the surfaces of the amorphous silicon film 304 and the quartz substrate 301 before the second formation of the amorphous silicon film.

  Further, another method may be used for forming the amorphous silicon films 305 and 306. For example, an amorphous silicon film can be formed by forming the amorphous silicon film as a whole to 75 nm by the above-described method, partially forming a mask, and partially reducing the film thickness by the above-described etching. .

  The amorphous silicon film 305 later becomes a semiconductor active layer of the memory element, and the amorphous silicon film 306 later becomes a semiconductor active layer of a switching element and a peripheral CMOS circuit.

  When the final thickness of the semiconductor active layer is 150 nm or more, particularly 200 nm or more, impact ionization peculiar to SOI is extremely small, which is almost the same as the case of non-volatile memory using bulk silicon. It will disappear. Therefore, the characteristics of the nonvolatile memory based on the SOI technology cannot be extracted. Therefore, in the present invention, the final thickness of the semiconductor active layer is preferably less than 150 nm (preferably less than 100 nm).

  Further, in this embodiment, as described above, the final film thickness of the amorphous silicon film 305 of the memory element is 50 nm, and the final film thickness of the amorphous silicon film 306 of the switching element and the peripheral CMOS circuit or the like. However, the thickness is preferably not limited to the film thickness of this embodiment, and may be in the range of 1 to 50 nm (more preferably 10 to 40 nm) and 40 to 100 nm.

It is important to thoroughly control the impurity concentration in the film when forming the amorphous silicon film. In this embodiment, the concentrations of C (carbon) and N (nitrogen) which are impurities that inhibit crystallization in the amorphous silicon films 305 and 306 are both less than 5 × 10 ¹⁸ atoms / cm ³ (typical 5 × 10 ¹⁷ atoms / cm ³ or less, preferably 2 × 10 ¹⁷ atoms / cm ³ or less), and O (oxygen) is less than 1.5 × 10 ¹⁹ atoms / cm ³ (typically 1 × 10 ¹⁸ atoms / cm ³ or less, preferably 5 × 10 ¹⁷ atoms / cm ³ or less). This is because the presence of each impurity at a concentration higher than this will adversely affect the subsequent crystallization and cause deterioration of the film quality after crystallization. In the present specification, the impurity element concentration in the film is defined by the minimum value in the measurement result of SIMS (mass secondary ion analysis).

In order to obtain the above-described configuration, it is desirable that the low-pressure thermal CVD furnace used in this embodiment periodically perform dry cleaning to clean the film formation chamber. In dry cleaning, 100 to 300 sccm of ClF ₃ (chlorine fluoride) gas is allowed to flow in a furnace heated to about 200 to 400 ° C., and the film formation chamber may be cleaned with fluorine generated by thermal decomposition.

  According to the knowledge of the present inventors, when the furnace temperature is 300 ° C. and the flow rate of ClF 3 (chlorine fluoride) gas is 300 sccm, 4 μm of deposits (mainly containing silicon as a main component) are 4 Can be completely removed in time.

  Further, the hydrogen concentration in the amorphous silicon films 305 and 306 is also a very important parameter, and it seems that a film with better crystallinity can be obtained by keeping the hydrogen content low. Therefore, the amorphous silicon films 305 and 306 are preferably formed by a low pressure thermal CVD method. Note that the plasma CVD method can be used by optimizing the film formation conditions.

  Next, a crystallization process of the amorphous silicon films 305 and 306 is performed. As a means for crystallization, the technique described in Japanese Patent Application Laid-Open No. 7-130552 by the present inventor is used. Either means of Example 1 or Example 2 of the publication can be used, but it is preferable to use the technical contents described in Example 2 (detailed in Japanese Patent Application Laid-Open No. 8-78329) in the present invention.

  In the technique described in Japanese Patent Application Laid-Open No. 8-78329, first, mask insulating films 307 to 309 for selecting the addition region of the catalyst element are formed. Then, a solution containing nickel (Ni) as a catalyst element for promoting crystallization of the amorphous silicon films 305 and 306 is applied by a spin coating method to form a Ni-containing layer 310 (FIG. 4A).

  In addition to nickel, the catalytic element is cobalt (Co), iron (Fe), palladium (Pd), platinum (Pt), copper (Cu), gold (Au), germanium (Ge), lead (Pb). Indium (In) or the like can be used.

  The catalyst element addition step is not limited to the spin coating method, and an ion implantation method or a plasma doping method using a resist mask can also be used. In this case, since the occupied area of the added region can be reduced and the growth distance of the lateral growth region can be easily controlled, this is an effective technique for configuring a miniaturized circuit.

  Next, when the catalyst element addition step is completed, after dehydrogenation at 450 ° C. for about 1 hour, a temperature of 500 to 700 ° C. (typically 550 to 650 ° C.) in an inert atmosphere, hydrogen atmosphere or oxygen atmosphere. The amorphous silicon films 305 and 306 are crystallized by applying a heat treatment for 4 to 24 hours. In this embodiment, heat treatment is performed at 570 ° C. for 14 hours in a nitrogen atmosphere.

  At this time, the crystallization of the amorphous silicon films 305 and 306 proceeds preferentially from the nuclei generated in the regions 311 and 312 to which nickel is added, and the crystal region 313 grown almost parallel to the substrate surface of the substrate 301. 314 and 315 are formed. The inventors refer to these crystal regions 313, 314, and 315 as lateral growth regions. Since the lateral growth regions are relatively aligned and individual crystals are gathered, there is an advantage that the overall crystallinity is excellent (FIG. 4B).

  After the heat treatment for crystallization is completed, the mask insulating films 307, 308, and 309 are removed and patterning is performed to form island-like semiconductor layers (active layers) 316 to 319 including only a lateral growth region.

  Next, the channel formation region of the island-shaped semiconductor active layer 316 and 317 to 319 are covered with resist masks 320 and 321, and impurity ions imparting P-type are added. In this embodiment, B (boron) is used as an impurity element, but In (indium) may be used. Note that the acceleration voltage at the time of impurity addition is about 80 keV.

  Therefore, the source / drain regions 125 and 127 of the island-shaped semiconductor active layer 316 and the channel formation region 126 are formed. Further, since the island-like semiconductor active layers 317 to 319 are covered with the resist mask 321, no impurities are added.

After that, the resist mask 321 is removed, and a gate insulating film 325 made of an insulating film containing silicon is formed (FIG. 5A). The thickness of the gate insulating film 325 may be adjusted within a range of 10 to 250 nm in consideration of an increase due to a later thermal oxidation process. The thickness of the gate insulating film of the memory element island-shaped semiconductor active layer may be 10 to 50 nm, and the thickness of the other gate insulating film may be 50 to 250 nm. For the gate insulating film, SiO ₂ , SiON, SiN or the like may be used. As a film formation method, a known vapor phase method (plasma CVD method, sputtering method, or the like) may be used.

  Next, as shown in FIG. 5A, heat treatment (catalyst element gettering process) for removing or reducing the catalyst element (nickel) is performed. In this heat treatment, a halogen element is included in the treatment atmosphere, and the gettering effect of the metal element by the halogen element is used.

  Note that the heat treatment is preferably performed at a temperature exceeding 700 ° C. in order to obtain a sufficient gettering effect by the halogen element. Below this temperature, decomposition of the halogen compound in the processing atmosphere becomes difficult, and the gettering effect may not be obtained. Therefore, the heat treatment temperature is preferably 800 to 1000 ° C. (typically 950 ° C.), and the treatment time is 0.1 to 6 hr, typically 0.5 to 1 hr. It is necessary to prevent impurities existing in the source / drain regions from diffusing into the channel region during the heating.

  As a typical example, in an atmosphere containing hydrogen chloride (HCl) at a concentration of 0.5 to 10% by volume (in this example, 3% by volume) with respect to an oxygen atmosphere, 950 ° C., 30 minutes. The heat treatment may be performed. If the HCl concentration is equal to or higher than the above concentration, the surface of the active layers 316 to 319 is not preferable because irregularities of about the film thickness occur.

In addition to the HCl gas, the compound containing a halogen element is one or more selected from compounds containing a halogen element such as HF, NF ₃ , HBr, Cl ₂ , ClF ₃ , BCl ₃ , F ₂ , Br _2, etc. A seed can be used.

In this step, nickel in the active layers 316 to 319 is gettered by the action of chlorine and becomes volatile nickel chloride which is separated into the atmosphere and removed. By this step, the concentration of nickel in the active layers 316 to 319 is reduced to 5 × 10 ¹⁷ atoms / cm ³ or less (typically 2 × 10 ¹⁷ atoms / cm ³ or less). According to the experience of the present inventors, if the nickel concentration is 1 × 10 ¹⁸ atoms / cm ³ or less (preferably 5 × 10 ¹⁷ atoms / cm ³ or less), the TFT characteristics are not adversely affected.

The gettering process is also effective for metal elements other than nickel. As the metal elements that can be mixed into the silicon film, the constituent elements of the deposition chamber (typically aluminum, iron, chromium, etc.) can be considered, but if the gettering process is performed, the concentration of these metal elements is considered. 5 × 10 ¹⁷ atoms / cm ³ or less (preferably 2 × 10 ¹⁷ atoms / cm ³ or less).

When the gettering process is performed, the halogen elements used for the gettering process remain in the active layers 316 to 319 at a concentration of 1 × 10 ¹⁶ to 1 × 10 ²⁰ atoms / cm ³ .

  In addition, a thermal oxidation reaction proceeds at the interface between the active layers 316 to 319 and the gate insulating film 325 by the heat treatment, and the thickness of the gate insulating film 325 increases by the amount of the thermal oxide film. When the thermal oxide film is formed in this way, a semiconductor / insulating film interface with very few interface states can be obtained. In addition, there is an effect of preventing formation defects (edge thinning) of the thermal oxide film at the end portion of the active layer.

  Furthermore, it is also effective to improve the film quality of the gate insulating film 325 by performing heat treatment at 950 ° C. for about 1 hour in a nitrogen atmosphere after the heat treatment in the halogen atmosphere.

  Next, a metal film mainly composed of aluminum (not shown) is formed, and later gate electrode prototypes 129 to 132 are formed by patterning. In this embodiment, an aluminum film containing 2 wt% scandium is used. In addition, a tantalum film, a conductive silicon film, or the like can also be used (FIG. 5B).

  Here, the technique described in Japanese Patent Laid-Open No. 7-135318 by the present inventors is used. This publication discloses a technique for forming source / drain regions and low-concentration impurity regions in a self-aligning manner using an oxide film formed by anodic oxidation. The technique will be briefly described below.

  First, anodization is performed in a 3% oxalic acid aqueous solution while leaving a resist mask (not shown) used for patterning the aluminum film, thereby forming porous anodic oxide films 330 to 337. Since this film thickness later becomes the length of the low concentration impurity region, the film thickness is controlled accordingly.

  Next, after removing a resist mask (not shown), anodization is performed in an electrolytic solution in which 3% tartaric acid is mixed with an ethylene glycol solution. In this process, dense non-porous anodic oxide films 338 to 341 are formed. The film thickness may be 70 to 120 nm.

  Then, the aluminum films 342 to 345 remaining after the above-described two anodic oxidation treatments substantially function as gate electrodes (FIG. 5C). Note that the aluminum film 342 later becomes a floating gate electrode of the memory element.

  Next, the gate insulating film 325 is etched by dry etching using the gate electrodes 342 to 345 and the porous anodic oxide films 330 to 337 as a mask, and patterned to 346 to 349 (FIG. 5D).

  Then, the porous anodic oxide films 330 to 337 are removed (FIG. 6A). The end portions of the gate insulating films 346 to 349 thus formed are exposed by the thickness of the porous anodic oxide films 330 to 337.

  Next, the gate electrode 342 is divided to form a floating gate electrode 342 '(FIG. 6B).

  Next, an impurity element adding step for imparting one conductivity is performed. As the impurity element, P (phosphorus) or As (arsenic) may be used for N type, and B (boron) or In (indium) may be used for P type.

First, resist masks 350 and 351 are formed in order to add impurities to the N-type TFT. In this embodiment, the impurity addition is performed in two steps. First, the first impurity addition (P (phosphorus) is used in this embodiment) is performed at a high acceleration voltage of about 80 keV to form an n ⁻ region. The n ⁻ region is adjusted so that the P ion concentration is 1 × 10 ¹⁷ atoms / cm ³ to 1 × 10 ¹⁹ atoms / cm ³ .

Further, the second impurity addition is performed at a low acceleration voltage of about 10 keV to form an n ⁺ region. At this time, since the acceleration voltage is low, the gate insulating film functions as a mask. The n ⁺ region is adjusted so that the sheet resistance is 500Ω or less (preferably 300Ω or less).

  Therefore, source / drain regions 352 to 355 of the N-type TFT, low-concentration impurity regions 356 and 357, and channel regions 358 and 359 are formed.

Next, as shown in FIG. 6D, resist masks 360 and 361 are provided so as to cover the N-type TFT, and impurity ions imparting P-type (boron is used in this embodiment) are added, and p ⁻ A region, p ⁺ region, is formed. The p ⁻ region is adjusted so that the boron ion concentration is 1 × 10 ¹⁷ atoms / cm ³ or more (preferably 1 × 10 ¹⁸ atoms / cm ³ or more). In addition to boron, Ga, In, or the like may be used.

  Thus, source / drain regions 362 and 363, a low concentration impurity region 364, and a channel formation region 367 of the P-type TFT are formed (FIG. 6D).

  As described above, since the low-concentration impurity regions are formed in the switching TFT and other peripheral circuit TFTs, impact ionization is less likely to occur even when the semiconductor active layer is thin.

  When the active layer is completed as described above, the impurity element is activated by a combination of furnace annealing, laser annealing, lamp annealing and the like. At the same time, the damage of the active layer received in the addition process is also repaired.

  Note that it can be said that the channel formation region of the TFT of this embodiment is substantially single crystal with a small number of mismatch bonds.

  Next, an interlayer insulating film 368 is formed to a thickness of 500 nm. As the interlayer insulating film 368, a silicon oxide film, a silicon nitride film, a silicon oxynitride film, an organic resin film, or a stacked film thereof can be used.

  Next, after forming contact holes, source / drain electrodes 369 to 374 and a control gate electrode 375 of the memory element are formed. The control gate electrode 375 is connected to the upper surface of the anodic oxide film 338 (FIG. 7B).

  Finally, the entire substrate is heated in a hydrogen atmosphere at 350 ° C. for 1 to 2 hours, and the entire device is hydrogenated to terminate dangling bonds (unpaired bonds) in the film (particularly in the active layer). Through the above steps, a TFT having a structure as shown in FIG. 7B can be manufactured.

(Knowledge about impurities contained in active layer)
The active layer (semiconductor thin film) of this embodiment is characterized in that C (carbon), N (nitrogen), and O (oxygen), which are elements that inhibit crystallization, are not present or substantially absent. This is a configuration that can be achieved through thorough impurity (contamination) management.

In the case of the present embodiment, since the mixing of C (carbon), N (nitrogen) and O (oxygen) is thoroughly avoided in the formation of the amorphous silicon film, it inevitably exists in the final semiconductor film. The concentration of C (carbon) and N (nitrogen) is at least less than 5 × 10 ¹⁸ atoms / cm ³ (typically 5 × 10 ¹⁷ atoms / cm ³ or less, preferably 2 × 10 ¹⁷ atoms / cm ³ or less), The concentration of O (oxygen) is at least less than 1.5 × 10 ¹⁹ atoms / cm ³ (typically 1 × 10 ¹⁸ atoms / cm ³ or less, preferably 5 × 10 ¹⁷ atoms / cm ³ or less).

Incidentally, since the concentration of silicon is about 5 × 10 ²² atoms / cm ³ in a semiconductor film made of pure silicon, for example, an impurity element of 5 × 10 ¹⁸ atoms / cm ³ exists at a concentration of about 0.01 atomic%. It corresponds to doing.

  In addition, it is desirable that the concentration of C (carbon), N (nitrogen), and O (oxygen) present in the final semiconductor film is less than the lower limit of detection in SIMS analysis, and more preferably not completely present. It is necessary to obtain high crystallinity.

  As a result of analysis by the SIMS by the present inventors, when an amorphous silicon film in which the concentrations of C, N, and O all satisfy the above concentration range is used as a starting film, C contained in the active layer of the completed TFT It has been found that the concentrations of N, O also satisfy the above concentration range.

  Here, FIG. 8A shows a circuit layout of the nonvolatile memory of this embodiment. 8B is a cross-sectional view taken along the line A-A ′ in FIG. 8A, and FIG. 8C is an equivalent circuit diagram of FIG.

  In FIG. 8A, reference numerals 801 to 808 denote semiconductor active layers, which constitute TFTs Tr1 to Tr8. Reference numerals 809 to 812 denote first wiring layers, which are used as gate electrodes of Tr2, Tr4, Tr6, and Tr8, wiring of gate signal lines, and wiring of gate signal lines of Tr1, Tr3, Tr5, and Tr7. Note that the floating gate electrodes 813 to 816 of Tr1, Tr3, Tr5, and Tr7 are formed at the same time as the first wiring layer and are in a floating state after patterning. Reference numerals 817 to 828 denote second wiring layers which are used as signal lines connected to the source / drain regions of the respective Trs and as control gate electrodes of the Tr1, Tr3, Tr5 and Tr7. In the drawing, the blacked out portion indicates that the lower wiring or the semiconductor layer is in contact. In the figure, all wirings of the same pattern are the same wiring layer.

  In the nonvolatile memory of the present invention, the thickness of the semiconductor active layer of the memory element is made thinner than the thickness of the semiconductor active layer of the element constituting the switching element or other peripheral circuit. Impact ionization is likely to occur, and the memory element can be written / erased at a low voltage. This leads to a decrease in deterioration with respect to the number of times of writing / erasing of the memory element. This is an innovative solution to the problem that the gate insulating film has deteriorated because the gate insulating film is relatively thin in the conventional EEPROM manufactured from bulk silicon. Further, in the conventional bulk silicon, the gate insulating film is thin, so that the carrier accumulated in the floating gate electrode flows out due to the temperature rise.

  In this embodiment, an inexpensive low-grade quartz substrate is first prepared. Next, the quartz substrate is polished to an ideal state (average value of the uneven portion difference is within 5 nm, typically within 3 nm, preferably within 2 nm) by a technique such as CMP (Chemical Mechanical Polishing).

  Thus, even an inexpensive quartz substrate can be used as an insulating substrate having excellent flatness by polishing. When a quartz substrate is used, the substrate becomes very dense and the stability of the substrate / semiconductor thin film interface is high. Further, since there is almost no influence of contamination from the substrate, the utility value is very high.

  In Example 1, an example in which a halogen element is used in the process of gettering a catalytic element that promotes crystallization of silicon is shown. In the present invention, it is also possible to use a phosphorus element in the catalyst element gettering step. Other steps shall be in accordance with Example 1.

In the case of using phosphorus element, phosphorus is added to a region other than the region to be the active layer, and heated at a temperature of 400 to 1050 ° C. (preferably 600 to 750 ° C.) for 1 min to 20 hr (typically 30 min to 3 hr). What is necessary is just to process. Since the catalytic element is gettered to the region where phosphorus is added by this heat treatment, the concentration of the catalytic element in the active layer is reduced to 5 × 10 ¹⁷ atoms / cm ³ or less.

  When the gettering process is completed in this manner, an active layer is formed using a region other than the region to which phosphorus is added. Thereafter, according to the steps of the first embodiment, a semiconductor device having the same characteristics as the first embodiment is obtained.

  Of course, if a heat treatment is performed in an atmosphere containing a halogen element when forming a thermal oxide film to be a gate insulating film, there is a synergistic effect between the gettering effect by the phosphorus element and the gettering effect by the halogen element in this embodiment. can get.

  In this embodiment, a case where an EEPROM is constituted by an inverted stagger type TFT will be described with reference to FIGS. In FIGS. 9 to 11, attention is paid only to one memory cell (memory element and switching element) constituting the EEPROM, but an address decoder, a peripheral circuit, and the like can be formed at the same time. Actually, as shown in FIG. 1 described in the first embodiment, an EEPROM is constituted by a plurality of memory cells arranged in a matrix, an address decoder, and peripheral circuits.

  Please refer to FIG. First, a base film 902 made of a silicon oxide film is provided on a glass substrate 901, and gate electrodes 903 and 904 are formed thereon. The gate electrode 903 later becomes a control gate electrode of the memory element, and the gate electrode 904 later becomes a gate electrode of the switching element. In this embodiment, a chromium film having a thickness of 200 nm to 400 nm is used as the gate electrodes 903 and 904, but an aluminum alloy, tantalum, tungsten, molybdenum, a silicon film imparted with conductivity, or the like may be used.

  Next, a gate insulating film 905 is formed to a thickness of 100 to 200 nm on the gate electrodes 903 and 904. As the gate insulating film 905, a silicon oxide film, a silicon nitride film, or a stacked film of a silicon oxide film and a silicon nitride film is used. An anodized film obtained by anodizing a gate electrode can also be used as a gate insulating film.

  The gate insulating film on the memory element side defines the capacitance between the floating gate electrode and the control gate electrode to be formed in the next step, and is applied to the floating gate electrode while changing its film thickness. The voltage can be adjusted. Therefore, the thickness of the gate insulating film 905 is not limited to the above range, and the thickness may be partially changed.

  Next, a floating gate electrode 906 is formed (FIG. 9B). In this embodiment, a chromium film is used as the floating gate electrode. However, an aluminum alloy, tantalum, tungsten, molybdenum, a silicon film imparted with conductivity, or the like may be used.

  Next, an insulating film 907 is formed to a thickness of 10 to 50 nm. As the insulating film 907, a silicon oxide film, a silicon nitride film, or a stacked film of a silicon oxide film and a silicon nitride film is used.

  Next, amorphous silicon films 908 and 909 are formed by the method shown in FIGS. 3A to 3D of Example 1 (FIG. 9C). In this embodiment, the final film thickness of the amorphous silicon film 908 of the memory element is 50 nm and the final film thickness of the amorphous silicon film 909 of the switching element is 75 nm. (Preferably 10 to 40 nm), it may be formed in the range of 40 to 100 nm, and is not limited to the film thickness of this embodiment. Although not shown, the amorphous silicon film of the TFT of the address decoder or peripheral circuit can be formed to have the same thickness as that of the switching element.

  Next, the amorphous silicon films 908 and 909 are irradiated with laser light or strong light having the same intensity as the laser light to crystallize the amorphous silicon film (FIG. 9D). As the laser light, excimer laser light is preferable. As the excimer laser, a pulse laser using KrF, ArF, or XeCl as a light source may be used.

  As strong light having the same intensity as laser light, strong light from a halogen lamp or a metal halide lamp, or strong light from an infrared light or ultraviolet light lamp can be used.

In this embodiment, excimer laser light processed into a linear shape is scanned from one end to the other end of the substrate to crystallize the entire surface of the amorphous silicon film. At this time, the sweep speed of the laser beam is 1.2 mm / s, the processing temperature is room temperature, the pulse frequency is 30 Hz, and the laser energy is 300 to 315 mJ / cm ² . A crystalline silicon film is obtained by this process.

  The crystallization method used in Example 1 or Example 3 can also be used for the amorphous silicon film of this example.

  In addition, it is understood that the crystallization method of this example can be used for the amorphous silicon film of Example 1.

  Reference is now made to FIG. The crystalline silicon film is patterned to form active layers 910 and 911.

Next, an impurity element imparting one conductivity is added. First, an active layer for forming channel regions of the memory element, the N-type TFT, and the P-type TFT is covered with a resist mask (not shown), and an impurity element imparting P-type (boron is used in this embodiment). added may also be) using, p boron ion concentration of 1 × 10 ¹⁷ atoms / cm ³ or more (preferably 1 × 10 ¹⁸ atoms / cm ³ or more) ^- region (low concentration impurity regions, not shown ).

Next, resist masks 912 and 913 are formed (FIG. 10B). Then, an impurity element imparting P-type is added so as to have a concentration of about 1 × 10 ¹⁸ to 1 × 10 ²⁰ atoms / cm ³ to form a source region 914 and a drain region 915 of the P-type TFT. In addition, a portion of the active layer covered with the resist mask 912 becomes a channel region (FIG. 10B).

Next, the resist masks 912 and 913 are removed, and resist masks 917 and 918 are formed. Then, an impurity element imparting N-type (in this embodiment, phosphorus is used. Arsenic or the like may be used) is added, and a low concentration of about 1 × 10 ^{17 to} 5 × 10 ¹⁸ atoms / cm ³ is added. Impurity regions 919 and 920 are formed (FIG. 10C).

Next, the resist masks 917 and 918 are removed, and resist masks 921 and 922 are formed. Then, an impurity element imparting N-type is added again at a higher concentration (1 × 10 ¹⁸ to 1 × 10 ²⁰ atoms / cm ³ ) than in the step of FIG. 10C, and source / drain regions 923 of the N-type TFT are added. And 924 are formed. Note that reference numeral 925 denotes a low concentration impurity region, and reference numeral 926 denotes a channel formation region (FIG. 10D).

  Next, after removing the resist masks 921 and 922, excimer laser light is irradiated (laser annealing) to recover damage during ion implantation and activate the added impurities (FIG. 11A).

  When the laser annealing is finished, an interlayer insulating film 927 is formed to a thickness of 300 to 500 nm (FIG. 11B). The interlayer insulating film 927 is formed of a silicon oxide film, a silicon nitride film, an organic resin, or a laminated film thereof.

  Next, contact holes are formed in the interlayer insulating film 927, and source / drain electrodes 928, 929, and 930 made of a metal thin film are formed. As this metal thin film, aluminum, tantalum, titanium, tungsten, molybdenum, or a stacked film thereof may be used (FIG. 11B).

  Next, the whole is subjected to heat treatment at 350 ° C. for about 2 hours in a hydrogen atmosphere, and dangling bonds in the film (particularly the channel formation region) are terminated with hydrogen. The state shown in FIG. 11B is obtained through the above steps.

  The non-volatile memories of Examples 1 to 4 have various uses. In this embodiment, a semiconductor device using these nonvolatile memories will be described.

  Examples of such a semiconductor device include a video camera, a still camera, a head mounted display, a car navigation, a personal computer, a portable information terminal (mobile computer, mobile phone, etc.), and the like. An example of them is shown in FIG.

  FIG. 12A illustrates a mobile phone, which includes a main body 1201, an audio output unit 1203, an audio input unit 1203, a display device 1204, operation switches 1205, and an antenna 1206. The nonvolatile memory of the present invention may be integrally formed with the display device 1204.

  FIG. 12B illustrates a video camera, which includes a main body 1301, a display device 1302, an audio input portion 1303, operation switches 1304, a battery 1305, and an image receiving portion 1306. The nonvolatile memory of the present invention may be formed integrally with the display device 1302.

  FIG. 12C illustrates a mobile computer, which includes a main body 1401, a camera unit 1402, an image receiving unit 1403, operation switches 1404, and a display device 1405. The nonvolatile memory of the present invention may be formed integrally with the display device 1405.

  FIG. 12D illustrates a head mounted display which includes a main body 1501, a display device 1502, and a band portion 1503. The nonvolatile memory of the present invention may be formed integrally with the display device 1502.

  In this embodiment, a case where Ta (tantalum) or a Ta alloy is used for the gate electrode in the manufacturing method described in the first to fifth embodiments will be described.

When Ta or Ta alloy is used for the gate electrode, it can be thermally oxidized at about 450 ° C. to about 600 ° C., and an oxide film with good film quality such as Ta ₂ O ₃ is formed on the gate electrode. This oxide film is known to have better film quality than the oxide film formed when Al (aluminum) is used as the gate electrode described in the first embodiment.

  This was found by the fact that the oxide film of Ta or Ta alloy has better characteristics than the oxide film of Al in the JE characteristic (current density-electric field strength characteristic), which is one of the breakdown voltage evaluations of the insulating film. .

Ta ₂ O ₃ has a relative dielectric constant of around 11.6 and has a large capacitance between the floating gate and the control gate, so that charges are injected into the floating gate as compared with the case where Al is used for the gate electrode. There is also an advantage that it is easy.

  Further, when Ta is used for the gate electrode, it can be anodized as in the above embodiment.

  (CGS knowledge)

  Here, a semiconductor thin film manufactured by the manufacturing method described in Example 1 will be described. According to the manufacturing method of Example 1, the amorphous silicon film can be crystallized to obtain a crystalline silicon film called continuous grain boundary crystalline silicon (so-called Continuous Grain Silicon: CGS).

  The laterally grown region of the semiconductor thin film obtained by the manufacturing method of Example 1 has a unique crystal structure composed of an aggregate of rod-like or flat rod-like crystals. The characteristics are shown below.

  [Knowledge about the crystal structure of the active layer]

  When viewed microscopically, the lateral growth region formed in accordance with the manufacturing process of Example 1 has a crystal structure in which a plurality of rod-like (or flat rod-like) crystals are arranged in parallel in a specific direction in a substantially parallel manner. This can be easily confirmed by observation with a TEM (transmission electron microscope).

  In addition, the present inventors have observed the crystal grain boundaries of the semiconductor thin film obtained by the above-described manufacturing method by 8 million times using HR-TEM (High Resolution Transmission Electron Microscopy) and observed in detail (FIG. 13 (A)). However, in this specification, a crystal grain boundary is defined as a grain boundary formed at a boundary where different rod-shaped crystals are in contact with each other unless otherwise specified. Therefore, for example, it is considered to be distinguished from a grain boundary in a macro sense where different lateral growth regions collide with each other.

  By the way, the above-mentioned HR-TEM (High Resolution Transmission Electron Microscopy) is a method in which an electron beam is irradiated perpendicularly to a sample and the atomic / molecular arrangement is evaluated using interference of transmitted electrons and elastically scattered electrons. It is. By using this method, it is possible to observe the arrangement state of crystal lattices as lattice fringes. Therefore, by observing the crystal grain boundary, it is possible to infer the bonding state between atoms at the crystal grain boundary.

  In the TEM photograph (FIG. 13A) obtained by the present inventors, it was clearly observed that two different crystal grains (rod-like crystal grains) were in contact with each other at the crystal grain boundary. Further, at this time, it has been confirmed by electron beam diffraction that the two crystal grains have a roughly {110} orientation, although the crystal axis includes some deviation.

  By the way, in the lattice stripe observation by the TEM photograph as described above, the lattice stripe corresponding to the {111} plane was observed in the {110} plane. Note that the lattice stripe corresponding to the {111} plane refers to a lattice stripe in which a {111} plane appears in a cross section when crystal grains are cut along the lattice stripe. It can be simply confirmed by the distance between the lattice fringes which surface the lattice fringes correspond to.

  At this time, the present inventors obtained a very interesting finding as a result of observing in detail a TEM photograph of the semiconductor thin film obtained by the production method of Example 1 described above. In two different crystal grains that can be seen in the photograph, lattice fringes corresponding to the {111} plane were seen in both. And it was observed that the plaids of each other were clearly running in parallel.

  Further, regardless of the existence of the crystal grain boundary, lattice fringes of two different crystal grains are connected so as to cross the crystal grain boundary. In other words, it was confirmed that most of the lattice fringes observed across the crystal grain boundary are linearly continuous despite the fact that they are lattice fringes of different crystal grains. This is the same at an arbitrary crystal grain boundary, and 90% or more (typically 95% or more) of the lattice fringes keeps continuity at the crystal grain boundary.

  Such a crystal structure (exactly, the structure of the crystal grain boundary) indicates that two different crystal grains are joined with extremely good consistency at the crystal grain boundary. That is, the crystal lattice is continuously connected at the crystal grain boundary, and the trap level caused by crystal defects or the like is very difficult to create. In other words, it can be said that the crystal lattice has continuity at the grain boundaries.

  In FIG. 13B, the present inventors also analyzed a conventional polycrystalline silicon film (so-called high-temperature polysilicon film) by electron beam diffraction and HR-TEM observation as a reference. As a result, the lattice fringes of the two different crystal grains ran completely apart, and there was almost no joining that continued with good consistency at the grain boundaries. That is, it was found that there are many portions where lattice fringes are interrupted (portions indicated by arrows, etc.) at the crystal grain boundary, and there are many crystal defects. In such a portion, a dangling bond is present, and there is a high possibility of inhibiting the movement of carriers as a trap level.

  The present inventors call the bonding state of atoms when lattice fringes correspond with good matching like the semiconductor thin film obtained by the manufacturing method of Example 1 described above as matching bonding, and the bonding hand at that time is the matching bonding hand. Call. Conversely, as is often seen in conventional polycrystalline silicon films, the bonding state of atoms when lattice fringes do not correspond with good matching is called mismatch bonding, and the bond at that time is the mismatch bond (or unpaired bond). Called hand).

  Since the semiconductor thin film used in the nonvolatile memory of the present invention has extremely excellent matching at the crystal grain boundary, the above-mentioned mismatched bonds are very few. As a result of the inventors' investigation of a plurality of crystal grain boundaries, the proportion of mismatched bonds to the total bonds is 10% or less (preferably 5% or less, more preferably 3% or less). . That is, 90% or more (preferably 95% or more, more preferably 97% or more) of the total bonds are constituted by matched bonds.

  In addition, FIG. 14A shows the result of observation of the lateral growth region manufactured according to the manufacturing method of Example 1 described above by electron beam diffraction. FIG. 14B shows an electron diffraction pattern of a conventional polysilicon film (what is called a high temperature polysilicon film) observed for comparison.

  14 (A) and 14 (B) are measured with the diameter of the electron beam irradiation spot being 1.35 μm, it is considered that information of a sufficiently macro region is picked up compared to the lattice fringe level. Good.

  FIG. 14C is a schematic diagram of an electron beam diffraction pattern when an electron beam is irradiated perpendicularly to the {110} plane of single crystal silicon. Usually, such an electron diffraction pattern is compared with the observation result to infer what the orientation of the observation sample is.

  In the case of FIG. 14A, diffraction spots corresponding to <110> incidence appear relatively cleanly as shown in FIG. 14C, and the crystal axis is the <110> axis (the crystal plane is {110} Can be confirmed.

  Each spot has a slightly concentric spread, which is presumed to have a certain rotation angle distribution around the crystal axis. The extent of the spread is within 5 ° even if estimated from the pattern.

  In addition, during many observations, there were cases where the diffraction spots were not partially visible (a part of the diffraction spots could not be seen even in FIG. 14A). Although it is probably {110} orientation, the diffraction pattern seems to be invisible because the crystal axis is slightly shifted.

  Based on the fact that the {111} plane is almost always included in the crystal plane, the present inventors presume that such a phenomenon is probably caused by a shift in the rotation angle around the <111> axis.

  On the other hand, in the case of the electron beam diffraction pattern shown in FIG. 14B, it is confirmed that the diffraction spots do not have clear regularity and are oriented almost randomly. That is, it is expected that crystals having a plane orientation other than the {110} plane are irregularly mixed.

  As these results show, the crystalline silicon film produced by the manufacturing method of Example 1 described above is characterized in that almost all crystal grains are roughly oriented in the {110} plane and continuous to the lattice at the crystal grain boundaries. It is to have sex. This feature is not present in conventional polysilicon films.

  As described above, the semiconductor thin film manufactured by the manufacturing method of Example 1 described above was a semiconductor thin film having a crystal structure (exactly, the structure of a crystal grain boundary) completely different from that of the conventional semiconductor thin film. The present inventors also described the results of analysis of the semiconductor thin film used in the present invention in Japanese Patent Application Nos. 9-55633, 9-165216, and 9-212428.

  Note that the present inventors performed X-ray diffraction according to the method described in Japanese Patent Laid-Open No. 7-321339, and calculated the orientation ratio of the crystalline silicon film of the above-described manufacturing method. In this publication, the orientation ratio is defined by a calculation method as shown in the following Equation 1.

  FIG. 17 shows an example of the result of measuring the orientation of the semiconductor thin film described above by X-ray diffraction. In the X-ray diffraction pattern, a peak corresponding to the (220) plane appears, but it goes without saying that it is equivalent to the {110} plane. As a result of this measurement, it was found that the {110} plane was the main orientation, and the orientation ratio was 0.7 or more (typically 0.9 or more).

  As described above, it can be seen that the crystalline silicon film produced by the manufacturing method of Example 1 described above and the conventional polysilicon film have completely different crystal structures (crystal structures). Also from this point, it can be said that the crystalline silicon film of the present invention is a completely new semiconductor film.

  In forming the semiconductor thin film of Example 1 described above, the annealing process at a temperature higher than the crystallization temperature plays an important role in reducing defects in crystal grains. Explain that.

  FIG. 15A is a TEM photograph in which the crystalline silicon film is enlarged by 250,000 times at the time when the crystallization process is completed in the manufacturing method of Example 1 described above. Appears due to a difference in contrast), and a defect that looks like a zigzag like the arrow is confirmed.

  Such defects are mainly stacking faults in which the stacking order of atoms on the silicon crystal lattice plane is different, but there are also cases such as dislocations. FIG. 15A seems to be a stacking fault having a defect plane parallel to the {111} plane. This can be inferred from the fact that the zigzag defect is bent at an angle of about 70 °.

  On the other hand, as shown in FIG. 15B, the crystal silicon film produced by the above-described manufacturing method of Example 1 viewed at the same magnification shows almost no defects due to stacking faults or dislocations in the crystal grains. It can be confirmed that the crystallinity is very high. This tendency is true for the entire film surface, and it is difficult to reduce the number of defects to zero, but it can be reduced to a level that can be regarded as substantially zero.

  That is, in the crystalline silicon film shown in FIG. 15B, defects in the crystal grains are reduced to an almost negligible level, and the crystal grain boundaries cannot be a barrier for carrier movement due to high continuity. It can be regarded as a crystal or substantially a single crystal.

  As described above, the crystal silicon films shown in the photographs of FIGS. 15A and 15B have substantially the same continuity in the crystal grain boundaries, but there is a large difference in the number of defects in the crystal grains. There is. The reason why the crystalline silicon film according to the manufacturing method of Example 1 described above exhibits much higher electrical characteristics than the crystalline silicon film shown in FIG. 15A is largely due to the difference in the number of defects.

  The crystal silicon film (FIG. 15B) obtained in this way by the manufacturing method of the above-described first embodiment has a crystal grain significantly higher than that of a crystal silicon film (FIG. 15A) that has just been crystallized. The number of defects is small.

The difference in the number of defects appears as a difference in spin density by electron spin resonance analysis (Electron Spin Resonance: ESR). At present, it has been found that the spin density of the crystalline silicon film by the manufacturing method of Example 1 described above is at least 5 × 10 ¹⁷ spins / cm ³ or less (preferably 3 × 10 ¹⁷ spins / cm ³ or less). However, since this measured value is close to the detection limit of existing measuring devices, the actual spin density is expected to be even lower.

  The present applicant calls the crystalline silicon film having the above-described crystal structure and characteristics as continuous grain boundary crystalline silicon (Continuous Grain Silicon: CGS).

  In the conventional semiconductor thin film, the crystal grain boundary functions as a barrier that prevents the movement of carriers. However, in the semiconductor thin film according to the manufacturing method of Example 1 described above, such a crystal grain boundary does not substantially exist, so that high carriers are present. Mobility is realized. Therefore, the electrical characteristics of the TFT manufactured using the semiconductor thin film by the manufacturing method of Example 1 described above show very excellent values. This is shown below.

  [Knowledge about electrical characteristics of TFT]

  Since the semiconductor thin film by the manufacturing method of Example 1 described above can be substantially regarded as a single crystal (substantially no crystal grain boundary exists), a TFT using it as an active layer is comparable to a MOSFET using single crystal silicon. Electrical characteristics are shown. The following data has been obtained from the TFT fabricated by the present inventors.

(1) Sub-threshold coefficient, which is an index of TFT switching performance (agility of switching on / off operation), is 60 to 100 mV / decade (typically 60 to 85 mV) for both N-channel and P-channel TFTs. / decade) and small.
(2) Field-effect mobility (μ _FE ), which is an indicator of TFT operating speed, is 200 to 650 cm ² / Vs (typically 250 to 300 cm ² / Vs) for N-channel TFTs, and P-channel TFTs 100 to 300 cm ² / Vs (typically 150 to 200 cm ² / Vs).
(3) The threshold voltage (V _th ), which serves as an index of TFT driving voltage, is as low as −0.5 to 1.5 V for N-channel TFTs and −1.5 to 0.5 V for P-channel TFTs.

  As described above, it has been confirmed that extremely excellent switching characteristics and high-speed operation characteristics can be realized.

  Note that the annealing process at a temperature (700 to 1100 ° C.) equal to or higher than the crystallization temperature described above plays an important role in reducing defects in crystal grains when forming CGS. This will be described below.

  From the above, it can be seen that the gettering process of the catalytic element is an indispensable step in producing the CGS. The present inventors consider the following model for the phenomenon that occurs in this process.

  First, in the state shown in FIG. 15A, a catalyst element (typically nickel) is segregated in defects (mainly stacking faults) in crystal grains. That is, it is considered that there are many bonds in the form of Si-Ni-Si.

  However, if Ni present in the defect is removed by performing the gettering process of the catalytic element, the Si-Ni bond is broken. For this reason, the surplus bonds of silicon immediately form a Si-Si bond and become stable. Thus, the defect disappears.

  Of course, it is known that defects in the crystalline silicon film disappear due to thermal annealing at a high temperature, but the bond with nickel is broken and the recombination of silicon is smooth because many bonds are generated. I can guess it will be done.

  In addition, the inventors say that the heat treatment is performed at a temperature equal to or higher than the crystallization temperature (700 to 1100 ° C.) so that the crystalline silicon film and the base are fixed, and the defects are eliminated by increasing the adhesion. I am also thinking about a model.

[Knowledge about the relationship between TFT characteristics and CGS]
The excellent TFT characteristics as described above are largely due to the use of a semiconductor thin film having continuity in the crystal lattice at the crystal grain boundary as the active layer of the TFT. The reason is discussed below.

  The continuity of the crystal lattice at the crystal grain boundary results from the fact that the crystal grain boundary is a grain boundary called “planar grain boundary”. The definition of the planar grain boundary in this specification is “Characterization of High-Efficiency Cast-Si Solar Cell Wafers by MBIC Measurement; Ryuichi Shimokawa and Yutaka Hayashi, Japanese Journal of Applied Physics vol.27, No.5, pp.751”. -758, 1988 ”is the“ Planar boundary ”.

  According to the above paper, the planar grain boundaries include {111} twin grain boundaries, {111} stacking faults, {221} twin grain boundaries, {221} twist grain boundaries, and the like. This planar grain boundary is characterized by being electrically inactive. That is, although it is a crystal grain boundary, it does not function as a trap that inhibits the movement of carriers, and thus can be regarded as substantially nonexistent.

  In particular, the {111} twin boundaries are also called Σ3 corresponding grain boundaries, and the {221} twin boundaries are also called Σ9 corresponding grain boundaries. The Σ value is a parameter that serves as a guideline indicating the degree of consistency of the corresponding grain boundary. It is known that the smaller the Σ value, the better the grain boundary.

  As a result of observing the semiconductor thin film by the manufacturing method of Example 1 described above in detail by TEM, most of the crystal grain boundaries (90% or more, typically 95% or more) are Σ3 corresponding grain boundaries. That is, it was found to be a {111} twin grain boundary.

  In the crystal grain boundary formed between two crystal grains, when the plane orientation of both crystals is {110}, assuming that the angle formed by the lattice stripes corresponding to the {111} plane is θ, θ = 70.5 ° It is known that sometimes it becomes the corresponding grain boundary of Σ3.

  Therefore, in the crystal grain boundary shown in the TEM photograph of FIG. 13A, each lattice fringe of adjacent crystal grains is continuous at an angle of about 70 °, and this crystal grain boundary is a {111} twin grain boundary. It can be easily inferred that

  Incidentally, when θ = 38.9 °, the corresponding grain boundary of Σ9 is obtained, but such other crystal grain boundaries also existed.

  Such a corresponding grain boundary is formed only between crystal grains having the same plane orientation. That is, the semiconductor thin film produced by the manufacturing method of Example 1 described above can form such a corresponding grain boundary over a wide range because the plane orientation is approximately {110}. This feature is not possible with other polysilicon films with irregular surface orientation.

  Here, FIG. 16A shows a TEM photograph (dark field image) obtained by enlarging the semiconductor thin film by 15,000 times according to the manufacturing method of Example 1 described above. There are areas that appear white and areas that appear black, but the portions that appear the same color indicate that the orientation is the same.

  It should be noted in FIG. 16A that the white-looking region is continuously gathered at a considerable rate in such a wide dark field image. This means that crystal grains having the same orientation exist with a certain degree of orientation, and adjacent crystal grains have almost the same orientation.

  On the other hand, FIG. 16B shows a TEM photograph (dark field image) obtained by enlarging a conventional high-temperature polysilicon film by 15,000 times. In the conventional high-temperature polysilicon film, the portions having the same plane orientation are scattered in a scattered manner, and it is not possible to confirm a group having a direction as shown in FIG. This is considered because the orientation of adjacent crystal grains is completely irregular.

  In addition to the measurement points shown in FIG. 13, the present inventors have repeated observation and measurement over a large number of regions, and in a wide region sufficient to produce a TFT, the crystal lattice continues in the crystal grain boundary. It is confirmed that the sex is maintained.

  Further, FIG. 18 shows a TEM photograph in the case where the semiconductor thin film is observed in a bright field when the nickel gettering process is performed using phosphorus in the manufacturing method of Example 3 described above. In FIG. 18, a photograph of Point 1 enlarged 300,000 times is shown in FIG. 19A, and a photograph enlarged 2 million times is shown in FIG. 19B. Note that a region surrounded by a square in FIG. 19A corresponds to FIG. An electron diffraction pattern (spot diameter 1.7 μmφ) at Point 1 is shown in FIG.

  Furthermore, Point 2 and Point 3 were observed under exactly the same conditions as Point 1. The observation results of Point 2 are shown in FIGS. 20 (A), 20 (B), and 20 (C), and the observation results of Point 3 are shown in FIGS. 21 (A), 21 (B), and 21 (C). .

  From these observation results, it can be seen that continuity is maintained in the crystal lattice at an arbitrary crystal grain boundary, and a planar grain boundary is formed. In addition to the measurement points shown here, the present inventors have repeated observation and measurement over a large number of regions, and the continuity of the crystal lattice at the crystal grain boundary is large enough to produce a TFT. It is confirmed that it is secured.

It is a circuit diagram of the nonvolatile memory of the present invention. It is sectional drawing of the memory element and switching element which comprise the non-volatile memory of this invention. It is a figure which shows the manufacturing process of the non-volatile memory of this invention. It is a figure which shows the manufacturing process of the non-volatile memory of this invention. It is a figure which shows the manufacturing process of the non-volatile memory of this invention. It is a figure which shows the manufacturing process of the non-volatile memory of this invention. It is a figure which shows the manufacturing process of the non-volatile memory of this invention. FIG. 3 is a top view, a cross-sectional view, and a circuit diagram of a memory element and a switching element that constitute the nonvolatile memory of the present invention. It is a figure which shows the manufacturing process of the non-volatile memory of this invention. It is a figure which shows the manufacturing process of the non-volatile memory of this invention. It is a figure which shows the manufacturing process of the non-volatile memory of this invention. It is a figure showing an example of a semiconductor device using a nonvolatile memory of the present invention. It is a TEM photograph figure which shows the crystal grain of a semiconductor thin film. It is a photograph figure which shows the electron diffraction pattern of a semiconductor thin film. It is a TEM photograph figure which shows the crystal grain of a semiconductor thin film. It is a TEM photograph figure which shows the dark field image of a semiconductor thin film. It is a graph which shows the result of the X-ray diffraction of a semiconductor thin film. It is a TEM photograph figure which shows the dark field image of a semiconductor thin film. It is the TEM photograph figure and electron diffraction pattern figure which show the crystal grain boundary of a semiconductor thin film. It is the TEM photograph figure and electron diffraction pattern figure which show the crystal grain boundary of a semiconductor thin film. It is the TEM photograph figure and electron diffraction pattern figure which show the crystal grain boundary of a semiconductor thin film.

Explanation of symbols

101 X address decoder 102 Y address decoder 201 substrate 202 semiconductor active layer 203, 204 source / drain region 205 channel forming region 206 semiconductor active layer 207, 208 source / drain region 209 low concentration impurity region 210 channel forming region 211, 212 gate insulation Film 213 Floating gate electrode 214 Anodized film 215 Control gate electrode 216, 219, 220 Source / drain electrode 217 Gate electrode 218 Anodized film 221 Interlayer insulating film

Claims (1)

A semiconductor device having a nonvolatile memory,
The nonvolatile memory includes a plurality of memory cells in which memory elements and switching elements are arranged in a matrix.
The memory element includes a semiconductor active layer provided on an insulating substrate, a gate insulating film, a floating gate electrode, an oxide film provided to cover the floating gate electrode, and a control gate electrode. ,
The switching element includes a semiconductor active layer provided on the insulating substrate, a gate insulating film, and a gate electrode,
The semiconductor device according to claim 1, wherein the thickness of the semiconductor active layer of the memory element is smaller than the thickness of the semiconductor active layer of the switching element.