Static semiconductor memory cell formed in an n-well and p-well

[0002] 1. Field of the Invention

[0003] The present invention relates to a static semiconductor memory device (hereinafter simply referred to as an SRAM).

[0004] 2. Description of the Background Art

[0005] Generally, a memory cell for an SRAM is formed of six elements in total including four n type transistors (Q1, Q2: access transistors, Q3, Q4: driver transistors) and two p type transistors (Q5, Q6: load transistors), as shown in FIG. 11. Two storage nodes 19aand 19bwhich are cross coupled have a bistable state of (H, L) or (L, H) and do not change their states if a prescribed power supply voltage is applied.

[0006] For data writing, a flip-flop state is set by selecting a word line to open gates (transfer gates) of access transistors Q1 and Q2 and forcing a voltage to be applied to a pair of bit lines in accordance with a desired logic value. For data reading, the above mentioned transfer gates are opened and the potentials of storage nodes 19aand 19bare transmitted to the bit lines. It is noted that, in FIG. 11, a cell current 20 is shown which flows from the Low side of storage nodes 19aand 19bof the memory cell to a ground line (a GND line) through a bit line BL or a complementary bit line /BL from a bit line load (not shown-in the drawing) during reading operation.

[0007] FIG. 12 shows a layout of a memory cell for an SRAM of the type disclosed in Japanese Patent Laying-Open No. 8-186181, for example. It is noted that a power supply line, ground line, bit line and the like are not shown for the convenience of the drawing.

[0008] Referring to FIG. 12, a memory cell 1 has n and p wells 2 and 3 provided adjacent to each other. Load transistors Q5 and Q6 are formed in n well 2. Access transistors Q1, Q2 and driver transistors Q3, Q4 are formed in p well 3.

[0009] A pair of word lines 17aand 17bare provided over memory cell 1, and a gate of driver transistor Q3 is connected to p and n type impurity regions through contacts 18aand 18b, respectively. In addition, a gate of driver transistor Q4 is connected to p and n type impurity regions through contacts 18cand 18d, respectively.

[0010] As shown in FIG. 12, n and p wells 2 and 3 are provided adjacent to each other in a direction in which word lines 17aand 17bextend, making memory cell 1 longer in the direction of the word lines. Thus, a pitch of a metal interconnection which functions as a bit line or the like increases. In addition, a capacitance between metal interconnections is reduced so an SRAM capable of operating at a high speed is obtained.

[0011] However, memory cell 1 is longer in the direction of word lines 17aand 17bas described above, resulting in longer word lines 17aand 17bwhen such memory cells are arranged in a matrix. Consequently, there arises a problem associated with a signal delay caused by a word line (hereinafter referred to as a “word line delay”).

[0012] The present invention is made to solve the aforementioned problem. It is an object of the present invention to provide an SRAM which has a memory cell including transistor formation regions of different conductivity types provided in a direction of a word line and which is capable of preventing the word line delay.

[0013] According to one aspect of the present invention, an SRAM includes a memory cell, a word line and first and second transistor regions. The memory cell includes a pair of access transistors, a pair of driver transistors and a pair of load transistors. The word line is provided for the pair of access transistors. The pair of load transistors are formed in the first transistor region. The second transistor region is provided adjacent to the first transistor region in the direction of the word line and has the pair of access transistors and the pair of driver transistors.

[0014] In order to reduce a resistance of the word line to prevent the word line delay, the word line can be formed of metal. In the conventional example shown in FIG. 12, however, as two word lines are formed for a single memory cell and p and n wells are arranged in a direction in which the word lines extend, two metal interconnections must be formed in a direction of the shorter sides of the memory cell so as to form the word line of metal. As a result, a pitch between the metal interconnections is made small, whereby the formation of the metal interconnection becomes difficult and a capacitance between the metal interconnections increases. On the other hand, in the present invention, as only one word line is provided, the word line can easily be formed of metal and the resistance of the word line can be reduced. Thus, the word line delay can be prevented.

[0015] Preferably, the above mentioned word line is formed of metal. Thereby, the resistance of the word line can be reduced and the word line delay is prevented as described above.

[0016] In addition, the SRAM includes first and second memory cells which are arranged in the direction of the word line such that the second transistor regions are adjacent to each other. A metal ground line, which is shared by the first and second memory cells, is provided over the word line in a direction which is orthogonal to the word line. Pairs of metal bit lines for the first and second memory cells are arranged on opposite sides of the metal ground line.

[0017] As the metal ground line which are shared by the first and second memory cells is provided in the direction orthogonal to the word line as described above, a cell current for the two memory cells flows to the single metal ground line. Thus, the increase in the potential of the ground line due to the cell current can effectively be prevented.

[0018] In addition, a field shield separation (isolation) region may be formed in the memory cell. In this case, preferably, the field shield separation region between the pair of driver transistors is continuously formed in the direction which is orthogonal to the word line to traverse the memory cell.

[0019] By forming the field shield separation region as described above, generation of an isolated region in the field shield separation region can effectively be prevented when a plurality of memory cells are arranged in a matrix. When such isolated region is formed, a contact for fixing a potential of the isolated region must be formed, thereby causing a problem that the metal interconnection cannot freely be patterned. In the present invention, however, the metal interconnection can more freely be patterned as compared with the case where the isolated region exists since the isolated region is not generated as described above.

[0020] In addition, a first impurity region is shared by one access transistor and one driver transistor and a second impurity region is shared by the other access transistor and the other driver transistor. Preferably, the interval between the ones of the access and driver transistors differs from that between the others of the access and driver transistors.

[0021] As shown in FIG. 1, for example, a channel width of the driver transistor is generally set greater than that of the access transistor. In this case, if the intervals between the access and driver transistors are different as described above, the driver transistors can be formed offset in the direction which is orthogonal to the word line. Thereby, the memory cell can be reduced in length in the direction of the word line as compared with the case where the driver transistors are arranged spaced by an equal distance from the word line. This is also contributable to the prevention of the word line delay.

[0022] According to another aspect, an SRAM of the present invention includes a memory cell, a word line and first and second transistor regions. The memory cell includes a pair of access transistors, a pair of driver transistors and a pair of load transistors, each having a gate. The word line is provided over the memory cell. The pair of load transistors, are formed in the first transistor region. The second transistor region is provided adjacent to the first transistor region in the direction in which the word line extends and has the pair of access transistors and the pair of driver transistors. The gate of the load transistor is orthogonal to that of the driver transistor.

[0023] As the gate of the load transistor is orthogonal to that of the driver transistor as described above, one of the gates can be provided in the direction which is orthogonal to the word line. Thereby, the memory cell is reduced in length in the direction of the word line and expanded in the direction which is orthogonal to the word line. As a result, the word line can be reduced in length and the word line delay can be prevented. In addition, even when two word lines are provided as in the conventional example, the word lines can be formed of metal as the memory cell is increased in length in the direction which is orthogonal to the word line. In this case, the word line delay can more effectively be prevented.

[0024] The above mentioned memory cell is formed on a semiconductor layer provided on a substrate with an insulation film interposed. In other words, according to the present aspect, an SRAM has an SOI (Semiconductor On Insulator) structure. The above mentioned load transistor and driver transistor have pairs of first and second impurity regions of first and second conductivity types, respectively. In this case, preferably, ones of the first and second impurity regions are in contact with each other and the others of the first and second impurity regions are arranged orthogonal to each other.

[0025] As the first and second impurity regions are arranged orthogonal to each other as described above, the gates of the load and driver transistors are also arranged orthogonal to each other, whereby the aforementioned effect is obtained. In addition, when the SOI structure is employed as in the present aspect, the first and second impurity regions of different conductivity types can be abut against each other. Thus, the memory cell can be further reduced in length in the direction of the word line. This is also contributable to the prevention of the word line delay.

[0026] In addition, the pair of access transistors and the pair of driver transistors may be arranged in a single line in the direction which is orthogonal to the word line.

[0027] Thereby, the memory cell can be further reduced in length in the direction of the word line as compared with the case described above.

[0028] The 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 drawings.

[0029] FIG. 1 is a plan view showing a memory cell in a manufacturing process up to a first metal interconnection for an SRAM in accordance with a first embodiment of the present invention.

[0030] FIG. 2 is a plan view showing the memory cell for the SRAM in accordance with the first embodiment of the present invention.

[0031] FIG. 3 is a cross sectional view taken along the line III—III in FIGS. 1 and 2.

[0032] FIG. 4 is a plan view showing the memory cells in FIG. 1 arranged in a matrix.

[0033] FIG. 5 is a plan view showing a memory cell in a manufacturing process up to first metal interconnection in accordance with a modification of the first embodiment.

[0034] FIG. 6 is a plan view showing the memory cell in accordance with the modification of the first embodiment.

[0035] FIG. 7 is a cross sectional view taken along the line VII—VII in FIGS. 5 and 6.

[0036] FIG. 8 is a plan view showing a memory cell in a manufacturing process up to a first metal interconnection for an SRAM in accordance with a second embodiment of the present invention.

[0037] FIG. 9 is a plan view showing the memory cell for the SRAM in accordance with the second embodiment of the present invention.

[0038] FIG. 10 is a cross sectional view taken along the line X—X in FIGS. 8 and 9.

[0039] FIG. 11 is a diagram showing an equivalent circuit for the SRAM.

[0040] FIG. 12 is a plan view showing an example of a memory cell structure for a conventional SRAM.

[0041] Embodiments of the present invention will now be described with reference to FIGS. 1 to 10.

[0042] Referring first to FIGS. 1 to 7, a first embodiment of the present invention and a modification thereof will be described. FIG. 1 is a plan view showing a memory cell 1 for an SRAM in accordance with the first embodiment of the present invention. It is noted that a memory cell up to a first metal interconnection is shown in FIG. 1 for the convenience of the explanation. A second metal interconnection is shown in FIG. 2. FIG. 3 is a cross sectional view taken along the line III—III in FIGS. 1 and 2.

[0043] Referring first to FIGS. 1 and 3, memory cell 1 includes an n well (a first transistor region) 2 and a p well (a second transistor region) 3. Load transistors Q5 and Q6 are formed on n well 2. Load transistor Q5 includes a polysilicon gate 7dand p type impurity regions 6aand 6b. Load transistor Q6 includes a polysilicon gate 7eand p type impurity regions 6cand 6d.

[0044] Access transistors Q1, Q2 and driver transistors Q3, Q4 are formed on p well 3. Access transistor Q1 includes n type impurity regions 5dand 5eand a polysilicon gate 7a. Access transistor Q2 shares polysilicon gate 7awith access transistor Q1 and includes n type impurity regions 5aand 5b.

[0045] Driver transistor Q3 includes a polysilicon gate 7cand n type impurity regions 5fand 5e. Driver transistor Q4 includes a polysilicon gate 7band n type impurity regions 5band 5c.

[0046] A trench separation (isolation) region 4ais formed in the periphery of n type impurity regions 5ato 5fand a trench separation region 4bis formed in the periphery of p type impurity regions 6ato 6d. Trench separation regions 4aand 4bare formed by burying insulation films in trenches formed in a main surface of semiconductor substrate 12 as shown in FIG. 3.

[0047] Above the aforementioned polysilicon gates 7ato 7e, first metal interconnections 9ato 9jincluding aluminum, tungsten, copper or the like are formed with an interlayer insulation film 13 interposed as shown in FIG. 3. First metal interconnections 9aand 9bare connected to n type impurity regions 5aand 5dthrough contact holes 8aand 8d, respectively. First metal interconnection 9cfunctions as a word line and is connected to polysilicon gate 7athrough a contact hole 8g. First metal interconnection 9dis connected to n type impurity region 5b, polysilicon gate 7c, p type impurity region 6band polysilicon gate 7ethrough contact holes 8b, 8i, 8kand 8o, respectively.

[0048] First metal interconnection 9eis connected to n type impurity regions 5cand 5fthrough contact holes 8cand 8f, respectively. First metal interconnection 9fis connected to polysilicon gate 7b, n type impurity region 5eand p type impurity region 6cthrough contact holes 8h, 8eand 8l, respectively. First metal interconnections 9g, 9hand 9iare connected to p type impurity region 6a, polysilicon gate 7dand p type impurity region 6dthrough contact holes 8j, 8nand 8m, respectively.

[0049] Referring now to FIGS. 2 and 3, second metal interconnections 11ato 11eincluding aluminum, tungsten, copper or the like are formed on first metal interconnections 9ato 9iwith interlayer insulation film 13 interposed. Second metal interconnection 11afunctions as a ground line and is connected to first metal interconnection 9evia a through hole 10a. Second metal interconnections 11cand 11bfunction as a bit line (BL) and a /bit line (/BL) . Second metal interconnections 11cand 11bare connected to first metal interconnections 9band 9avia through holes 10cand 10b, respectively.

[0050] Second metal interconnection 11dfunctions as a power supply line (a Vcc line) and is connected to first metal interconnections 9gand 9ivia through holes 10dand 10g, respectively. Second metal interconnection 11eis connected to first metal interconnections 9hand 9fvia through holes 10eand 10f, respectively. A plug is formed in each of through holes 10ato 10g.

[0051] As only one word line (9c) is provided and the word line is formed of metal as described above, a resistance of the word line can be reduced as compared with the conventional example. Thereby, the word line delay is prevented.

[0052] In addition, as shown in FIG. 1, an interval between access transistor Q1 and driver transistor Q3 is set larger than that between access transistor Q2 and driver transistor Q4. Therefore, driver transistors Q3 and Q4 are arranged offset with respect to each other in the direction which is orthogonal to word line (9c). Thus, even when channel widths of driver transistor Q3 and Q4 are greater than those of access transistors Q1 and Q2, the length of memory cell 1 in the direction of the word line can be kept small. This is also contributable to the prevention of the word line delay.

[0053] Referring now to FIG. 4, an exemplary arrangement of memory cell 1 shown in FIG. 1 will be described. As shown in FIG. 4, memory cells 1 are arranged in a matrix and a ground line (11a) is shared by two memory cells 1 which are provided adjacent to each other in the direction of the word line (9c). In this case, a set of memory cells 1 which shares the ground line (11a) are arranged in the direction of the word line (9c) such that p wells 3 are adjacent to each other. A BL (11c) and a /BL (11b) are provided for each memory cell 1 on opposite sides of ground line (11a).

[0054] As shown in FIG. 4, the ground line (11a), the BL (11c) and the /BL (11b) are provided in the direction which is orthogonal to the word line (9c), so that cell current 20 for two memory cells 1 flows through the single ground line (11a). Thereby, the increase in a potential of the ground line due to cell current 20 can be prevented. It is noted that a bit line contact is denoted by a reference numeral 18 in FIG. 4.

[0055] Referring to FIGS. 5 to 7, a modification of the above mentioned first embodiment will be described. FIGS. 5 and 6 are plan views showing memory cell 1 for an SRAM according to a modification of the first embodiment, and FIG. 7 is a cross sectional view taken along the line VII—VII in FIGS. 5 and 6.

[0056] Referring first to FIG. 7, in the present modification, the SRAM has an SOI (Semiconductor On Insulator) structure and a field shield separation (isolation) is employed. More specifically, an SOI layer (a semiconductor layer) is formed on a substrate 15 with an insulation film 16 interposed and transistors are formed in the SOI layer. In addition, field shield separation regions 14aand 14bare provided having field shield gates which are formed on the semiconductor layer with the insulation film interposed. Ground and power supply potentials are applied to field shield separation regions 14aand 14bfor separation on the sides of n and p type transistors, respectively. Thus, field shield separation regions 14aand 14bare separated as shown in FIG. 5.

[0057] In addition, contact holes 8p, 8rand 8tare provided for supplying potentials for the SOI layer in the field shield separation regions. Contact holes 8qand 8sare provided for fixing the potentials of the field shield gates. In addition, a first metal interconnection 9eis provided over contact holes 8qand 8rand connected to the SOI layer and the field shield gates therethrough. First metal interconnections 9kand 9jare formed on contact holes 8tand 8s.

[0058] Referring now to FIG. 6, in the present modification, second metal interconnection 11dextends even over first metal interconnections 9jand 9k. Second metal interconnection 11dis connected to first metal interconnections 9kand 9jvia through holes 10iand 10h, respectively. Other parts of the structure are almost the same as those for the case shown in FIG. 2.

[0059] An effect similar to that of the above described first embodiment can be obtained also in the case of the present modification. As shown in FIG. 5, field shield separation region 14abetween driver transistors Q3 and Q4 is continuously formed to traverse memory cell 1 in the direction which is orthogonal to the word line (9c). Therefore, also in the case where a plurality of memory cells 1 are arranged in a matrix, generation of an isolated region within the field shield separation region can effectively be prevented. When the isolated region is generated in the field shield separation region, a contact hole for fixing a potential of the isolated region must be formed, whereby a metal interconnection or the like cannot freely be formed. However, in the present modification, the isolated region is not formed as described above, so that the metal interconnection or the like can freely be formed.

[0060] Referring now to FIGS. 8 to 10, a second embodiment of the present invention will be described. FIGS. 8 and 9 are plan views of a memory cell for an SRAM in accordance with the second embodiment of the present invention. FIG. 10 is a cross sectional view taken along the line X—X in FIGS. 8 and 9.

[0061] In the second embodiment, SOI and trench separation structures are employed as shown in FIG. 10. Gates of load transistors Q5 and Q6 are provided orthogonal to those of driver transistors Q3 and Q4. Thereby, the gates of load transistors Q5 and Q6 can be provided in the direction which is orthogonal to word lines (7a, 7b), whereby a memory cell 1 can be Referring now to FIG. 9, in the present second embodiment, second metal interconnections 11ato 11care formed and the number of second metal interconnections is decreased as compared with the case of the first embodiment. Thus, even when the length of memory cell 1 is reduced in the direction of the word lines (7a, 7d), second metal interconnections 11ato 11ccan relatively easily be formed. It is noted that a first metal interconnection 9ffunctions as a power supply line. Further, the principle of the present second embodiment is applicable to a memory cell for an SRAM with the wells formed as shown in FIG. 1.

[0062] As in the foregoing, according to the SRAM of the present invention, the word line delay is effectively be prevented. Therefore, the high performance SRAM is obtained.

[0063] Although the present invention has been described and illustrated in detail, it is clearly understood that the same is by way of illustration and example only and is not to be taken by way of limitation, the spirit and scope of the present invention being limited only by the terms of the appended claims. reduced in length in the direction of the word line. Thus, the word line is shortened and the word line delay can be prevented.

[0064] As shown in FIG. 8, access transistors Q1, Q2 and driver transistors Q3, Q4 are arranged in a single line in the direction which is orthogonal to the word lines (7a, 7d). This is also contributable to the reduction in length of memory cell 1 in the direction of the word lines (7a, 7b). In addition, p type impurity regions 6aand 6care abut against n type impurity regions 5band 5d, respectively. Thereby, memory cell 1 can be reduced in length in the direction of the word lines (7a, 7d) as compared with the case where the wells of different conductivity types are formed as shown in FIG. 1. This is also contributable to the reduction in length of the word lines (7a, 7d).

[0065] In addition, further reduction in the size of the memory cell can be achieved as compared with the case shown in FIG. 1. Memory cell 1 shown in FIG. 1 requires a well separation width of 0.61 μm in accordance with the 0.18 μm rule. Here, as the length of memory cell 1 in the direction which is orthogonal to the word line (9c) is about 1.5 μm, an extra area due to the well separation would be 0.6×1.5=0.9 μm². On the other hand, in the present second embodiment, the well separation width is not necessary and the corresponding reduction in the area of memory cell 1 can be achieved.

[0066] Further, as shown in FIG. 8, in the present second embodiment, the length of memory cell 1 in the direction which is orthogonal to the word lines (7a, 7d) is longer than that in the direction in which the word lines (7a, 7d) extend. Therefore, even when two word lines (7a, 7d) are provided for a single memory cell 1, these word lines can be formed of metal as in the case of the first embodiment. In this case, further prevention of the word line delay can be achieved.

[0067] As shown in FIG. 8, p type impurity regions 6aand 6bare arranged in the direction which is orthogonal to n type impurity regions 5band 5c, and p type impurity regions 6cand 6dare arranged in the direction which is orthogonal to n type impurity regions 5cand 5d. Thereby, gates of load transistors Q5 and Q6 can be arranged orthogonal to those of driver transistors Q3 and Q4 as described above.