Static random access type semiconductor memory device

0001The present invention generally relates to semiconductor memory devices and, more particularly, relates to cell structures in a static random access memory where integration density of memory cells can be increased.

0002Fig. 7 is an equivalent circuit diagram of one memory cell in a conventional static random access memory (hereinafter referred to as SRAM). This memory cell includes thin-film p-type MOS transistors as loads. A pair of driver (for driving) transistors Q₁ and Q₂ (n-type MOS transistors) are connected to a pair of load transistors Q₅ and Q₆ (p-type MOS transistors) to form a flipflop circuit. The sources 110 and 111 of the pair of load transistors Q₅ and Q₆ are connected to a power supply Vcc and the sources 112 and 113 of driver transistors Q₁ and Q₂ are connected to GND. A pair of access transistors Q₃ and Q₄ (n-type MOS transistors) are connected to storage nodes 114 and 115, respectively. A bit line 107 is connected to one source/drain of access transistor Q₃ and a bit line 108 is connected to one source/drain of access transistor Q₄. The gates of access transistors Q₃ and Q₄ are connected to a word line 109.

0003Figs. 8 to 10 are plan views of the structure of an SRAM, showing three stages in order from the bottom on the surface of the substrate, respectively. Fig. 11 is a cross-sectional view of the structure taken along the line A-A in Figs. 8 to 10. Referring to Figs. 7, 8 to 11, a pair of driver transistors Q₁ and Q₂ and a pair of access transistors Q₃ and Q₄ are formed on a main surface of a p-type silicon substrate 148 of the memory cell. Driver transistor Q₁ includes a pair of source/drain regions 121 and 122 and a gate electrode 125. Driver transistor Q₂ includes a pair of source/drain regions 118 and 117 and a gate electrode 126. Access transistor Q₃ includes a pair of source/drain regions 119 and 120 and a gate electrode 109. Access transistor Q₄ includes a pair of source/drain regions 116 and 117 and a gate electrode 109. These transistors are n-type MOS transistors having source/drain regions formed on the main surface of p-type silicon substrate 148. Gate electrode 126 of driver transistor Q₂ is connected to source/drain region 120 of access transistor Q₃ through a contact 128. Gate electrode 126 of driver transistor Q₂ is connected to source/drain region 121 of driver transistor Q₁ through a contact 129. Gate electrode 125 of driver transistor Q₁ is connected to source/drain region 117 of access transistor Q₄ and source/drain region 117 of driver transistor Q₂ through a contact 127. A gate electrode 130 of a load transistor Q₅ is connected to a source/drain region 137 of a load transistor Q₆ through a contact 139. A gate electrode 131 of load transistor Q₆ is connected to a source/drain region 134 of load transistor Q₅ through a contact 138.

0004A bit line 107 is connected to source/drain region 119 of access transistor Q₃ through a contact 146 and a bit line 108 is connected to source/drain region 116 of access transistor Q₄ through a contact 147.

0005As stated above, in the memory cell of the conventional SRAM, four n-type MOS transistors are arranged on the silicon substrate and p-type thin film transistors are provided as loads above them. A case where a p-type thin film transistor is used as a load of a memory cell in an SRAM has been described in IEDM 1990 Technical Digest pp. 477-480. Fig. 13 is a cross-sectional view of a typical structure of a thin film transistor used as load transistors Q₅ and Q₆. The thin film transistor has a channel region 142 and a pair of source/drain regions 141 and 143 formed in a semiconductor layer such as polycrystalline silicon and a gate electrode 140 provided opposite to channel region 142 with an insulating layer interposed therebetween. Fig. 14 is a diagram showing a current characteristic of the thin film transistor.

0006In such an SRAM, it is necessary to reduce an area occupied by each memory cell in order to increase the integration density of the memory cells. However, the conventional memory cell above had two problems to be described below.

0007The first problem is that it is difficult to reduce an element isolation region between transistors making up the memory cell. Fig. 12 is a diagram showing by a model a cross-section of the structure of a LOCOS film 124 (Fig. 11) for insulating and isolating transistors from each other in the memory cell shown in Fig. 11. In this LOCOS film 152 (Fig. 12), regions X called "bird's beak" are formed at its both ends, which expand to the region where elements are formed, so that an isolation width W becomes larger than its desired value. For this reason, the width of the isolation region cannot be reduced, so that reduction in the size of the memory cell cannot be achieved.

0008The second problem concerns a current handling capability ratio β of a driver transistor to an access transistor (= the current handling capability of the driver transistor/the current handling capability of the access transistor). If the current handling capability ratio β is small, data is destroyed when it is read out from a memory cell. This phenomenon will now be described below. Figs. 15(a) and (b) show two inverter circuits obtained by dividing the equivalent circuit of the memory cell shown in Fig. 7 in connection with the reading characteristic. In this case, load transistors Q₅ and Q₆ are not shown because the amount of the current flowing through these load transistors is little enough to be ignored compared with those of the access transistors and the driver transistors, so that is has no effect on the reading operation. The characteristic of reading from a memory cell is given from a change in voltage at one storage node obtained by fixing the bit line and the word line at Vcc and changing the gate voltage of the driver transistor (the voltage at the other storage node). Fig. 16(a) is a diagram showing the reading characteristic in a case where the current handling capability ratio β is large (about 3). The axis of abscissa represents a voltage at storage node 115 and the axis of ordinate represents a voltage at storage node 114. The curve α₁ represents the voltage change characteristic at storage node 114 in a case where the voltage at storage node 115 is changed. The curve γ₁ shows the voltage change characteristic at storage node 115 in a case where the voltage at storage node 114 is changed. The curves α₁ and γ₁ intersect each other at three points P₁, P₂ and P₃. At point P₃, storage node 114 has "High" data stored, and storage node 115 has "High" data stored at point P₁. Point P₂ is an unstable point and the condition at this point P₂ is not kept at the time of reading. In the figure, a region surrounded by a circle h is called "eye of a memory cell". As the current handling capability ratio β of the transistors is larger, the circle h becomes bigger and the reading operation is stabilized.

0009In order to reduce the size of a memory cell, the size of an access transistor or a driver transistor is reduced. The access transistor or the driver transistor is reduced in size, for example, by shortening the gate length. If the transistor width of the access transistor is reduced to 1µm or less, a so-called narrow channel effect becomes significant, so that a threshold voltage Vth of the access transistor is increased. Fig. 16(b) shows the voltage change characteristic at the storage node in a case where the threshold voltage Vth of the access transistor is increased. In Figs. 16(a) and (b), Vcc-θ or Vcc-θ' corresponds to the threshold voltage Vth of the access transistor. As shown in Fig. 16(b), if the threshold voltage of the access transistor is increased, the curves α₂ and γ₂ intersect each other at one point P₂ only and the so-called "eye of a memory cell" region disappears. As a result, the stable points of the voltage at each storage node disappear, and data stored in the memory cell is destroyed at the time of the reading operation. For these reasons, the access transistor cannot be reduced in size even though the size of the driver transistor can be reduced. If only the driver transistor is reduced in size, the current handling capability ratio β of the both transistors becomes small, making the reading operation unstable.

0010IEEE J. Solid-State Circuits, Vol. SG-20, No.1, Feb. 1985, pp. 178 - 201, (Mahli et al.) discloses different types of thin film transistors in SOI technique and their characteristics.

0011An object of the present invention is to reduce the size of a memory cell and stabilize the operation of reading out stored data in an SRAM.

0012This object is achieved by a semiconductor memory device according to claim 1.

0013Further developments of the invention are given in the subclaims.

0014The semiconductor memory device includes a memory cell including a pair of driver transistors of a first conductivity type and a pair of load transistors of a second conductivity type making up a flipflop circuit, and a pair of access transistors, all transistors being thin film transistors. The semiconductor memory device further includes a semiconductor substrate having a main surface, an insulating layer formed on the semiconductor substrate, a first group of said thin film transistors arranged on the insulating layer, an interlayer insulating layer covering the surface of the first group of thin film transistors and a second group of said thin film transistors arranged on the interlayer insulating layer. The first group of thin film transistors includes at least one transistor of the driver transistors, the load transistors and the access transistors. The second group of thin film transistors includes at least one transistor of the driver transistor, the load transistor and the access transistor excluding the transistor included in the first group of thin film transistors.

0015The six transistors making up a memory cell are thin film transistors. Element isolation is made by providing the interlayer insulating layer between the thin film transistors. Accordingly, an area of an element isolation region can be reduced by eliminating the conventional element isolation structure using a LOCOS film.

0016Therefore, a narrow channel effect can be restrained and stabilization of the operation of reading out stored data as well as reduction in size of a memory cell can be achieved by forming a thin film transistor as an access transistor.

0017The foregoing and other objects, features, aspects and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying figures.

0018Fig. 1 is a plan view of the structure of a memory cell in an SRAM according to an embodiment of the present invention.

0019Fig. 2 is a plan view showing the structure of an upper layer portion of the memory cell shown in Fig. 1.

0020Fig. 3 is a cross-sectional view showing a typical structure of a thin film transistor used in this invention.

0021Fig. 4 is a diagram showing an electrical characteristic of the thin film transistor shown in Fig. 3.

0022Fig. 5 is a structural cross-sectional view taken along the line B-B shown in Figs. 1 and 2.

0023Fig. 6 is a structural cross-sectional view taken along the line C-C in Figs. 1 and 2.

0024Fig. 7 is an equivalent circuit diagram of a memory cell in a conventional SRAM.

0025Fig. 8 is a structural plan view of a memory cell in a conventional SRAM.

0026Fig. 9 is a structural plan view of a still upper layer of the memory cell shown in Fig. 8.

0027Fig. 10 is a structural plan view of a still upper layer of the memory cell shown in Fig. 9.

0028Fig. 11 is a structural cross-sectional view taken along the line A-A in Figs. 8 to 10.

0029Fig. 12 is a cross-sectional view of the structure in the vicinity of a LOCOS film used in isolating elements in a conventional memory cell.

0030Fig. 13 is a cross-sectional view showing the structure of a typical cross section of a thin film transistor used as a conventional load transistor.

0031Fig. 14 is a diagram showing an electrical characteristic of the thin film transistor shown in Fig. 45.

0032Fig. 15 is an equivalent circuit diagram (a), (b) showing two inverter circuits obtained by dividing the flipflop circuit shown in Fig. 7.

0033Fig. 16 is a diagram (a), (b) showing the characteristic curve of reading out data of a conventional memory cell.

0034A memory cell according to an embodiment includes pairs of access transistors Q₃ and Q₄, driver transistors Q₁ and Q₂ and load transistors Q₅ and Q₆ which are all thin film transistors. An equivalent circuit of this memory cell is equal to that shown in Fig. 7.

0035The structure of the memory cell will now be described with reference to Fig. 7 and further to Figs. 1, 2, 5 and 6. A main surface of a silicon substrate 32 has an insulating layer 33a formed thereon. A pair of access transistors Q₃ and Q₄ and a pair of driver transistors Q₁ and Q₂ are arranged on the surface of insulating layer 33a. The four transistors Q₁ to Q₄ are n-type thin film transistors. Access transistor Q₃ includes a pair of source/drain regions 1 and 3 and a channel region 2 formed in a polycrystalline silicon layer and a gate electrode 10. Access transistor Q₄ includes a pair of source/drain regions 6 and 8 and a channel region 7 formed in a polycrystalline silicon layer and a gate electrode 10. Driver transistor Q₁ includes a pair of source/drain regions 3 and 5 and channel region 4 formed in the polycrystalline silicon layer and a gate electrode 11. Driver transistor Q₂ includes a pair of source/drain regions 8 and 5 and a channel region 9 formed in the polycrystalline silicon layer and a gate electrode 12. The surfaces of these four transistors Q₁ to Q₄ are covered with a first interlayer insulating layer 33b.

0036A pair of p-type load transistors Q₅ and Q₆ are formed on the surface of first interlayer insulating layer 33b. Load transistor Q₅ includes a pair of source/drain regions 13 and 15, a channel region 14 and a gate electrode 11. Load transistor Q₅ shares gate electrode 11 with driver transistor Q₁. Load transistor Q₆ includes a pair of source/drain regions 13 and 17, a channel region 16 and a gate electrode 12. Load transistor Q₆ shares gate electrode 12 with driver transistor Q₂. The surfaces of load transistors Q₅ and Q₆ are covered with a second interlayer insulating layer 33c.

0037A pair of bit lines 107 and 108 are formed on the surface of second interlayer insulating layer 13c. Bit line 107 is connected to source/drain region 1 of access transistor Q₃ through a contact 26. Bit line 108 is connected to source/drain region 6 of access transistor Q₄ through a contact 27. Source/drain region 15 of load transistor Q₅ is connected to source/drain region 3 shared by access transistor Q₃ and driver transistor Q₁ through a contact 20. A contact 23 connects source/drain region 15 of load transistor Q₅ to an interconnection layer 18. A contact 22 connects interconnection layer 18 to gate electrode 12 of load transistor Q₆. A contact 21 connects source/drain region 17 of load transistor Q₆ to gate electrode 11 of load transistor Q₅. A contact 19 connects gate electrode 11 of load transistor Q₅ to source/drain region 8 of driver transistor Q₂.

0038Fig. 3 is a diagram showing by a model the structure of a typical cross section of a thin film transistor used in the embodiment. Fig. 4 shows an electrical characteristic of the thin film transistor shown in Fig. 3.

0039As stated above, since all the six transistors in the memory cell are thin film transistors, it is possible to prevent an increase in the threshold voltage of the access transistor under the influence of the narrow channel effect. Accordingly, the sizes of the access transistor and the driver transistor can be determined so that the current handling capability ratio β of the driver transistor to the access transistor is large. As a result, a memory cell can be constructed in which a stable reading operation can be performed.

0040As stated above, in accordance with one aspect of the present invention, all the transistors making up the memory cell are thin film transistors and the transistors are insulated and isolated from each other without using a LOCOS film, so that miniaturization of the cell structure by reduction in the size of the isolation region can be realized.