The present application claims priority from Japanese Patent Application No. 2011-183794 filed on Aug. 25, 2011, the content of which is hereby incorporated by reference into this application.
The present invention relates to a phase change memory device in which digital data is recorded by changing a state of a recording film between an amorphous phase and a crystal phase, and particularly to a technique useful in application to a phase change memory device having a three-dimensional structure.
According to spread of digital equipments and development of digital contents such as Internet, importance of a storage device which stores digital data therein increasingly becomes high.
Various performances are required for the storage device, but important performances among them are that a capacity of recording data is large, that a data transmission rate showing a time required to read/write a fixed amount of data is high, that power required to record a fixed amount of data is low, and the like.
A device to which attention is paid as a digital storage device in recent years is a solid state device (hereinafter, called “SSD”) using a semiconductor memory, centered on a flash memory. Since page data recording to/reading from the SSD is possible, the SSD includes such a matter that the data transmission rate is high as one of merits thereof, and a SSD for personal utilization in a personal computer or the like is frequently used because of this merit.
Furthermore, the SSD is expected as a storage device for business use requiring a high data transmission rate. However, since cost per bit (hereinafter simply referred to as “bit cost”) is currently high, application of the SSD to a storage system for business use is not very popular. One of the reasons is that the semiconductor device is manufactured via many steps. And it is necessary to make a cell area per bit small to achieve a large capacity in the SSD, and micromachining technique is required to do this.
Therefore, in order to reduce the bit cost of the SSD, it is required to increase a recording capacity per chip while suppressing increase of manufacturing cost per chip.
One example is a 3D technology of a semiconductor device described in Japanese Patent Application Laid-open Publication 2008-160004 (Patent Document 1). In an ordinary 3D technology, devices each composed of one layer are manufactured along an X-Y direction of a semiconductor substrate in a two-dimensional manner, and they are stacked so as to have a three-dimensional structure by any method. In this method, however, though cell area per bit can be reduced, manufacturing cost per bit cannot be reduced.
On the other hand, in the method described in Patent Document 1, recording regions holding data and Si (silicon) channels introducing current into the recording regions are continuously formed along a vertical direction (Z direction) of a semiconductor substrate. Furthermore, a plurality of gate electrodes of MOS transistors for selecting one of the recording regions in the vertical direction is formed in a vertical direction. Furthermore, these structures are arranged along the X-Y direction of the semiconductor device in a two-dimensional manner. Selection from a plurality of structures arranged in the X-Y direction is performed by a bit line and a word in the same manner as the DRAM or the like. According to such a structure, the recording region is selected three-dimensionally so that a three-dimensional recording device is realized.
According to the 3D technology in the above-mentioned Patent Document 1, since the structure along the vertical direction (Z direction) of the semiconductor device can be manufactured by a collective process, and size reduction in the X-Y direction becomes unnecessary due to recording many bits in the vertical direction, it is possible to increase the recording capacity per chip while suppressing increase of manufacturing cost per chip.
In recent years, as one of alternative technologies of the above-mentioned flash memory, a phase change memory device has been proposed. This phase change memory device is obtained by applying, to a semiconductor memory, a principle of DVD-RAM, DVD-RW, and the like which is a rewritable optical disk which have been put in the market, and digital data is recorded by reversibly changing a state of a recording film, which is made of a phase change material such as chalcogenide, between an amorphous phase and a crystal phase.
In the case of the phase change memory device, Joule heat generated by causing current to flowing into the recording film is utilized for reversible change between the amorphous phase and the crystal phase. That is, when the recording film is changed from the crystal phase to the amorphous phase, the recording film is locally melted by causing high current to flow in the recording film. At this time, when a cooling rate of the melted region after stop of current supply is sufficiently fast, since kinetic energy of atoms before atoms form arrangement of minimal energy becomes small, an amorphous phase is formed in a local area within the recording film. On one hand, when the recording film is changed from the amorphous phase to the crystal phase, current smaller than the above is caused to flow in the recording film to heat the recording film up to a temperature at which the recording film is crystallized. Furthermore, when data is read out, a difference in electric resistivity between the amorphous phase and the crystal phase is detected. Incidentally, the details of the technology about the phase change memory device have been summarized, for example, in Patent Document 1.
An example where the 3D technology described in the above-mentioned Patent Document 1 has been applied to the phase change memory device is described in Japanese Patent Application Laid-open Publication 2010-165982 (Patent Document 2). The phase change memory device has a structure where a recording film made of phase change material and a Si channel layer is in contact with each other, and it is configured such that an inversion layer is locally formed in the Si channel layer by applying a voltage to the gate electrode of a MOS transistor so that current is caused to flow in the recording film locally.
The structure of the above-mentioned phase change memory device is shown in
In the device shown in
Here, for example, when a voltage is applied to the gate electrode 27a of the MOS transistor so that the resistance of an area which is indicated by the broken line in the Si channel layer 25 becomes larger than that of the interface layer 24 and the phase change recording film 23, current passes through a path indicated by an arrow. In this way, since current can be introduced into a local portion of the phase change recording film 23 so as to form an inversion layer, it is possible to select a bit to be recorded and read.
Furthermore, in the device shown in
That is, since the Si channel layer 25 is higher in thermal conductivity than the central portion dielectric film 22, the temperature in the phase change recording film. 23 becomes low on the side closer the Si channel layer 25 than the central portion dielectric film 22, so that enormous power is required in order to raise the temperature of the phase change recording film 23 on the side of the Si channel layer 25 to a high temperature. Therefore, in order to solve this problem, the interface layer 24 for suppressing excessive thermal diffusion from the phase change recording film 23 to the Si channel layer 25 is provided.
One of the problems in the phase change memory device having a three-dimensional structure as shown in
That is, as described above, since thermal conductivity of the Si channel layer is high, when heat generated at the recoding bit is introduced into the Si channel layer, the heat diffuses to the adjacent bit along the Si channel layer. Here, when it is assumed that the adjacent bit is put in a crystal phase and the phase change recording film is locally melted in order to change the recording bit into the amorphous phase, it is thought that heat generated at the recording bit may erase the data in the adjacent bit.
This matter was calculated by a computer simulation. In this calculation, it is assumed that, in the device shown in
Here, the temperature at the minimum temperature point A of the recording bit and the temperature at the maximum attainment temperature point B of the adjacent bit shown in
However, when a voltage pulse is applied to the phase change recording film 23 so as to achieve Tremin of 630° C., there is such a possibility that due to manufacturing variation of the device or the like, the temperature of the phase change recording film 23 does not rise sufficiently so that a region which is not melted but crystallized occurs in the phase change recording film 23. The manufacturing variation of the device depends on precision of the manufacturing process of the device, but it is thought that manufacturing error generally occurs in a range of 5 to 10%. It is desirable from the fact that Trecmin is set to about 700° C. As the result of simulation which was performed under this condition, it was calculated that Tnbrmax was 130° C.
Next, the problem caused by the value will be described. Here, it is assumed that the adjacent bit is in an amorphous phase, and the recording bit is changed to an amorphous phase. In this case, when data is recorded in the recording bit, the adjacent bit must be prevented from being crystallized. In order to consider this matter, a crystallization probability is examined from a theoretical standpoint.
A time dependency χ(t) of the crystallization probability is expressed as follows:
Here, “t” is time, χ(t) is crystallization probability after elapsing of time “t”, “n” is index regarding dimension of crystal growth, “ν” is frequency, “Ea” is activation energy, “kB” is Boltzmann constant, and “T” is temperature. The reason why “k” is made dependent on time in the expression (1) and “k” is made dependent on temperature in the expression (2) is because the temperature “T” depends on time and “k” is integrated regarding time in the expression (1).
The values of the parameters in the above expressions are reported in detail, for example, in non-patent document “Journal of Applied Physics, 89, pp. 3168 to 3176 (2001)”. According to this document, Ea=2.0±0.2 eV (electron volt), ν=(1.5±0.3)×1022s−1, and n=2.5 in the phase change recording film (Ge2Sb2Te5) considered here. In another document, values such as Ea=2.4 eV and Ea=2.9 eV were reported. The Ea largely becomes engaged in the crystallization probability, where such a tendency appears that the lower Ea, the easier crystallization becomes. Considering that Ea becomes lower depending on the film formation condition of the phase change recording film, Ea=2.0 eV which is the minimum value in a plurality of documents is adopted.
Furthermore, the length of the recording bit in the Z direction and the length of the gate electrode in the Z direction are set to 30 nm (nanometers), respectively. These values depends on the recording capacity targeted by the device, and the smaller the values, the larger the thermal disturbance becomes, but when the value becomes smaller than 30 nm, it becomes difficult for the MOS transistor to form an inversion layer in the Si channel layer, and results in a thought that the minimum value is 30 nm.
Furthermore, the distance between the Si channel layer and the gate electrode is set to 10 nm. When the distance between the both is smaller than 10 nm, an insulation performance of the insulation film becomes lower, and when the distance becomes larger than 10 nm, electric field formed by the gate electrode spreads in the Si channel layer, so that the length of the inversion layer formed in the Si channel layer becomes long.
It is considered using the above expressions that how much Tnbrmax should be set when Trecmin=700° C. is set. Here, in the device shown in
The calculation result is shown in
Regarding χ, the value equal to 10−4 required as the maximum value is adopted as an error rate before error correction when Reed-Solomon code which is a method for error correction of digital data is used so that χ≦10−4 is set. Regarding Nrec, 106 times is sufficient as the number of times of rewriting for a device for personal use, but it is considered that 1012 times is required as the number of times of rewriting for a device for business SSD.
Considering Tnbrmax=130° C. which is the result of the above computer simulation together with the result shown in
An object of the present invention is to provide a technique for suppressing thermal disturbance of a phase change memory device having a three-dimensional structure.
Another object of the present invention is to provide a technique of increasing a recording capacity per chip while suppressing increase of bit cost of a phase change memory device having a three-dimensional structure.
The above-mentioned objects and other objects and novel features of the present invention will become apparent from the description of the specification and the accompanying drawings.
An outline of a representative one of the inventions disclosed in this application will be briefly described in the following manner.
The above problem is solved by applying a material having a high thermal conductivity to a gate insulation film of a MOS transistor for selection in a phase change memory device having a three-dimensional structure.
Here, in the case where the thermal conductivity of the gate insulation film 16 is κdiel, a temperature T at the minimum temperature point of the recording bit is 700° C., and a distance between the Si channel layer 15 and the gate electrode 17 of the selection MOS transistor, namely, an effective film thickness ddiel of the gate insulation film 16 is 10 nm, a simulation result regarding a relationship between a thermal conductivity κdiel of the gate insulation film 16 and a maximum attainment temperature point Tnbrmax of an adjacent bit is shown in
According to this result, Tnbrmax≦80° C. is found at Kdiel≧5. This fact can be easily understood from considering such a point that heat generated at the phase change recording film 13 is introduced into the Si channel layer 15 but the heat is introduced into the gate electrode 17 with a high thermal conductivity via the gate insulation film 16 so that the heat is sufficiently diffused. Furthermore, by considering the mechanism, it can also be understood that the thermal conductivity of the peripheral dielectric film 18 hardly affects the value of Tnbrmax. Therefore, the material of the peripheral dielectric film 18 may be the same as that of the gate insulation film 16, or the former may be different from the latter. Furthermore, the material of the gate electrode 17 hardly affects the value of Tnbrmax. This is because heat which has reached the gate electrode 17 diffuses sufficiently since a material having a high thermal conductivity, such as W or Si is used as the material of the gate electrode 17.
As described above, it is shown that the thermal conductivity of the gate insulation film 16 is required to be 5 W/m·K or more, but since the thermal conductivity of the gate insulation film material (SiO2) adopted in Patent Document 1 or Patent Document 2 is about 1.5 to 3 W/m·K, this value is insufficient to satisfy this requirement. For example, BN, Al2O3, AlN, TiO2, Si3N4, ZnO, and the like are proposed as an insulating material satisfying this requirement. And a mixture of these materials can be adopted. This is because the thermal conductivity of the mixture of the dielectrics shows an intermediate value of the thermal conductivities of the respective materials. Furthermore, in the case of the mixture of the dielectrics, a desirable thermal conductivity can be obtained by adjusting a mixing ratio of these materials.
In the above simulation, the film thickness ddiel of the gate insulation film 16 was set to 10 nm, but this value should be determined according to the electrical insulation property of the gate insulation film 16, and depends on the design of the device. Therefore, the thermal conductivity κdiel of the gate insulation film 16 should be determined using the value of this ddiel as a parameter.
The simulation result of the relationship between the film thickness ddiel of the gate insulation film 16 and the minimum thermal conductivity κdielmin required for the gate insulation film 16 is shown in
κdiel≧−9.36×10−3ddiel2+5.72×10−1ddeil−1.03×10−1 (3)
Here, the reason why κdielmin is subjected to fitting by a quadratic expression (3) is described qualitatively. First, one-dimensional thermal conduction is considered simply. In general, a heat source is a point, an initial temperature in a region other than the heat source is zero, and when heat diffuses isotropically, the temperature T(x, t) at the distance x after elapsing of time t can be expressed by the following expression (4).
This matter is described in detail, for example, in “Journal of Applied Physics, Vol. 66, pp. 1530-1543 (1989)” or the like.
Furthermore, the case of heat conduction to three dimensions, it is only necessary to substitute x2 in the expression (4) with (x2+y2+z2). Here, D is a constant including the thermal conductivity, and it is generally an amount called “thermal diffusivity”, but it can be thought that D is approximately proportional to the thermal conductivity. In the case considered in this invention, heat diffusing in the direction of the Si channel layer and heat diffusing in the direction of the gate electrode are sorted out from each other, and it is desirable that as much heat as possible diffuses in the direction of the gate electrode. That is, the temperature T (x, t) of the gate electrode is required to be raised to a certain value or more. When the value is represented as C, the following expression (5) is required by solving the expression (4).
√{square root over (D2 Ln(DC)≧ddiel)} (5)
From this expression (5), if Ln(DC) falls in an approximately constant range, ddiel and D, namely, κdielmin are approximately proportional to each other, but a deviation from the proportional relationship occurs due to existence of the logarithmic term. Fitting of the relationship between Ddiel and κdielmin is made possible by correcting an amount corresponding to the deviation by a quadratic expression. Incidentally, in the expression (3), the reason why the quadratic coefficient is small is because the deviation is slight.
In an actual MOS transistor, when the film thickness of the gate insulation film is thin, the insulation effect becomes small, and it encounters a problem that leakage occurs. Therefore, though application of a high-k material or the like is tried, the film thickness of the gate insulation film is generally 5 nm or more. That is, it is understood from the expression 3 and
Insulation film materials satisfying this demand are shown in Table 1. The thermal conductivity of the gate insulation film material (SiO2) used in the related art is typically about 1.4 W/m·K, and the gate insulation film material (SiO2) does not satisfy the demand of the present invention, while the other materials in Table 1 satisfy the demand of the present invention.
Incidentally, it is well known that, among dielectrics, AlN is a material showing the highest thermal conductivity, and that AlN easily mixes with oxygen, which results in reduction of the thermal conductivity of AlN due to phonon scattering caused by the mixing. It is described, for example, in “Journal of American Ceramic Society, 80, pp. 1421-1435 (1997)” that this problem can be solved by mixing Y2O3 into AlN at a rate of several percentages.
Furthermore, when an actual device is manufactured, adhesiveness of a stacked layer film is one of problems. It is known that adhesiveness between the Si channel layer and the gate insulation film (SiO2) in the related art is excellent, but when an alternative material for silicon oxide such as described in Table 1 is used, adhesiveness or mutual diffusion between elements may be problematic depending on manufacturing conditions. In particular, since the Si channel layer and the vicinity thereof reach a high temperature, there is a possibility that these problems become serious. These problems can be solved by providing a thin film made of SiO2 between the Si channel layer and the alternative material shown in Table 1.
That is, since SiO2 and the other alternative materials are the same dielectric, adhesiveness between these materials is excellent, and since SiO2 is a stable material, mutual diffusion of elements is also suppressed. As described above, however, since the thermal conductivity of SiO2 is low, heat conduction to the gate electrode is suppressed by providing the thin film made of SiO2.
Here, the simulation result of the relationship among the film thickness dSiO2 of SiO2, the film thickness ddiel of the gate insulation film including SiO2, and the required minimum value κdielmin of the thermal conductivity of the alternative material for silicon oxide is shown in
κ≧exp(BdSiO2) (6)
A=−9.3608×10−3 ddiel2+5.72039×10−1ddiel−1.03325×10−1 (7)
B=−2.04729×10−1 ln(ddiel)+1.33269 (8)
Here, as will be seen from the expression (6), an exponential function is used to dSiO2. The reason is because since the thermal conductivity of SiO2 is low, when dSiO2 increases, heat cannot be diffused sufficiently unless the thermal conductivity of the alternative material is made very large. Furthermore, in the case of dSiO2≧4 nm, the required minimum value Kdielmin satisfying the demand of the present invention can not be obtained within a range of 200 W/m·K or less. Therefore, dSiO2≦4 nm is demanded.
In the above explanation, attention is paid to thermal disturbance to the adjacent cell in Z direction. However, in the above configuration, there is such a concern that since heat is diffused to the gate electrode, thermal disturbance due to introduction to the adjacent cell in the X-Y direction through the gate electrode occurs. However, in the simulation result, it was confirmed that most temperature in a distance from one end of the gate electrode within 10 nm became close to the room temperature. Thereby, it is shown that even if heat is diffused to the gate electrode, the thermal disturbance to the adjacent cell in the X-Y direction is not problematic.
This result can be interpreted in the following manner. In the device structure shown in
An advantageous effect obtained by the inventions described in this application will be briefly described below.
In a phase change memory device with a three-dimensional structure which can record a large volume of digital data therein, the problem about the thermal disturbance where data is erased due to data rewriting in an adjacent bit can be solved.
That is, by adjusting the film thickness and the thermal conductivity of the gate electrode film in the selection MOS transistor, heat conducted along the Si channel layer can be diffused efficiently, and even if data rewriting is performed 1012 times, data erase can be suppressed sufficiently.
Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. Additionally, note that components having the same function are denoted by the same reference symbols throughout the drawings for explaining the embodiments, and the repetitive description thereof will be omitted. Furthermore, in the embodiments, explanation about the same or similar parts will not be repeated in principle. Furthermore, in the drawings for explaining the embodiments, hatching may be applied to even a plan view for easy understanding of configuration of this invention.
First of all, a manufacturing method of a phase change memory device shown in
Next, after the photoresist film 102 is removed, as shown in
Next, as shown in
Next, as shown in
Next, after the photoresist film 106 is removed, as shown in
For embedding these films into the through-hole 107, after an Al2O3 film having a film thickness of 10 nm, a Si film having a film thickness of 4 nm, a SiSb film having a film thickness of 4 nm, a Ge2Sb2Te5 film having a film thickness of 4 nm, and a SiO2 film having a film thickness of 8 nm are deposited on the peripheral dielectric film 105 in this order, for example, by the CVD method, these films left on the peripheral dielectric film 105 are polished and removed by the CMP method. Initial states of the SiSb film constituting the interface layer 111 and the Ge2Sb2Te5 film constituting the phase change recording film 110 are in crystal phase.
Next, as shown in
Next, as shown in
Next, as shown in
The word lines 116 are arranged so as to extend in a direction approximately perpendicular to the bit lines 101, and to extend approximately in the same direction as the gate electrodes 104. And by applying voltage to any of intersection points of the word lines 116 and the bit lines 101, the recording bit shown in
Here, electrode numbers (X1 to X5, Y1 to Y4, and Z1 to Z4) as shown in
In the phase change memory device of this embodiment, each state of the bits (X2, Y2, Z3) is firstly examined and recorded. In this embodiment, these bits are called “adjacent bits”.
Next, a current value obtained by applying DC voltages of 0.1 volts, 0 volts, and 3 volts to the respective electrodes X2, Y2, and Z3 was about 3.2 μA. That is, the resistance value is about 3.2 kΩ, and we think that the state of the recording film is in a crystal phase.
Next, a mark was recorded on the above recording bit. When voltage pulses of DC voltages of 3 volts, 5 volts, and 0 volts are applied to the respective electrodes Z3, X2, and Y2 with a time width of 30 ns, the resistance value reaches about 5 MΩ. From this, we think that the above-mentioned recording bit is changed to an amorphous state. This voltage pulse condition is hereinafter called “Reset pulse”. That is, the amorphous state is here defined as “Reset state”.
Furthermore, when voltage pulses of DC voltages of 3 volts, 3 volts, and 0 volts were applied to the electrode Z3, the electrode X2, and the electrode Y2 with a time width of 100 ns, respectively, the resistance value of the bit was changed to about 5 kΩ. From this, it was confirmed that change reversibly occurs between a crystal state and an amorphous state in this bit. The voltage pulse condition is hereinafter called “Set pulse”. That is, the crystal state is here defined as “Set state”. It was confirmed that the resistance value was changed to about 5 kΩ by applying the Reset pulse to this Set state again.
Next, the state of the bits (X2, Y2, Z2) adjacent to the adjacent bit in the Z direction was examined. In this case, these bits are called “recording bit”.
When examination was performed according to the same procedure as the above, we found that this bit was a crystal state. Next, when the resistance of the adjacent bit was examined by applying Reset pulse to this bit one time, the resistance value was about 5 MΩ which was the same as the original value. Thereafter, the result obtained by further applying Reset pulses to the recording bit continuously to measure the resistance value of the adjacent bit is shown in
From this result, decrease of the resistance value which is thought to be due to partial crystallization caused by pulse application of 1010 times appeared in the device of the related art, while the change of the resistance value did not appear even by pulse application of 1014 times in the device of the present invention, from which, it is found that the thermal disturbance is suppressed.
When the Reset pulse was applied to the adjacent bits of the device of the related art and the device of the present invention, the resistance values of the adjacent bits was changed to about 5 MΩ, and when the Set pulse was applied to it, the resistance values were changed to about 3 kΩ. From this, it was confirmed that a failure did not occur in the adjacent bits (X2, Y2, Z3).
Furthermore, when the set pulse was applied to the recording bit (X2, Y2, Z2), the resistance value was changed to 3.5 kΩ, and from this, it was confirmed that a failure did not occur in the recording bit. However, when Reset pulses were applied to the recording bit 1016 times, the resistance value of the recording bit was changed to 100 MΩ or more, and even when the Reset pulse voltage was then elevated, the resistance value was not changed at all. From this, it was considered that the recording bit was broken for any reason.
The same experiments as the above were performed regarding respective devices where the gate insulation films were made of TiO2, Al2O3, BN, Si3N4, and AlN, respectively, and as a result of these experiments, significant changes of the resistance values of the adjacent bits did not appear to the number of pulse application times of 1014 or less as well as the result shown in
In the second embodiment, as another example, the gate insulation film 16 of the device shown in
In this example, the film thicknesses of the above-mentioned SiO2 films are set to 1 nm, 2 nm, 3 nm, and 4 nm. Furthermore, other insulation films adjacent to the SiO2 film are made of TiO2, Al2O3, BN, Si3N4, and AlN, and their film thicknesses are set to 2 nm, 4 nm, 6 nm, 8 nm, and 10 nm. The interface layer and the phase change recording film are respectively composed of SiSb film and Ge2Sb2Te5 film as well as those of the first embodiment. Furthermore, a diameter of each through hole 107 (see
The same experiment as that in the first embodiment was performed using the above samples, and the resistance changes of the adjacent cells were measured after repetitive rewritings in a single cell.
The number of times of rewriting (the maximum number of times of rewriting) where the resistance in the adjacent cell decreased significantly is shown in
Incidentally, in this experiment, rewriting experiment of 1016 times or more was not performed. This is because it was determined that the fact that rewriting of 1016 times was possible sufficiently satisfied specifications as a storage device and as a memory.
Furthermore, in
From this experiment result, if it is assumed that the thermal conductivities of the insulation materials other than SiO2 are values described in Table 1, it is found that the expression (6) to (8) are satisfied. In
Though the invention which has been made by the present inventers has been specifically described above based upon the embodiments, the present invention is not limited to the embodiments, and it goes without saying that the present invention can be modified variously without departing from the gist of the present invention.
The present invention is available for a phase change memory device having a three-dimensional structure.
Number | Date | Country | Kind |
---|---|---|---|
2011-183794 | Aug 2011 | JP | national |