WO1984000244A1

WO1984000244A1 - Four-state rom cell with increased gain differential between states

Info

Publication number: WO1984000244A1
Application number: PCT/US1983/000643
Authority: WO
Inventors: Glenn Eric Noufer
Original assignee: Motorola Inc
Priority date: 1982-06-22
Filing date: 1983-05-02
Publication date: 1984-01-19
Also published as: EP0111534A1; EP0111534A4; US4546453A

Abstract

A four-state ROM cell (10) is improved by providing a tapered potential gate area (26) between a source region (30) and a drain region (28) which allows for the effective gate width to be increased and the gate length to be decreased for each succeedingly higher gain state with a single program mask at the polysilicon gate deposition stage.

Description

FOUR-STATE ROM CELL WITH INCREASED GAIN DIFFERENTIAL

BETWEEN STATES

Field of the Invention This invention relates to read only memory cells, and more particularly, to read only memory cells which can be programmed to more than the two states.

Background of the Invention In an attempt to make read only memory (ROM) arrays more dense, read only memory cells capable of being programmed to more than the two states typical for a ROM have been developed. As a practical matter and for convenience such cells are normally chosen to have four possible states although some other number of states can be chosen. For example, U.S. Patent No. 4,327,424 to

Philip T. Wu describes a three state cell. There are two primary approaches to developing a more-than-two-state transistor cell for MOS technology ROMs. One approach is to vary threshold voltages for the differing states. Because the threshold voltage of an MOS transistor is determined by the concentration of dopant in the channel, each additional state beyond the normal two states requires not only an additional process step for doping the selected channels but an additional program mask for masking the dopant from the unselected channels as well. The second approach has a transistor cell with a gain selected from more than two possible gains. The gain is directly related to the width to length ratio of the gate of the transistor. One of either the length or width is varied to achieve the desired variation in gain. U.S.

Patent No. 4,272,830 to Jerry D. Moench describes such an approach. Advantages are that no additional programming steps are required, and the one necessary programming step is relatively late in the process. A disadvantage is that in order to obtain sufficient gain differential for reliable detection of the state of the cell, the cell must be relatively large to accommodate the potential gate dimensions because only one of the length and width is varied.

Summary of the Invention An object of the invention is to provide an improved more-than-two-state ROM cell.

A further object is to provide a ROM cell which has a potential gate area of variable length and width.

Yet another object is to provide a ROM cell which has a potential gate area of variable length and width, both of which are selected by a single program mask.

And yet another object is to provide a transistor ROM cell with a potential gate area which is tapered toward one of the source and drain. These and other objects of the invention are carried out by a read only memory cell comprising a transistor having a first current electrode, a second current electrode, and a control electrode of a predetermined width and length, wherein the cell is capable of being programmed into more than two states as determined by the width and length of the control electrode, the control electrode having a potential area between the first and second current electrodes, the potential control electrode area characterized as being tapered between the first and second current electrodes.

Brief Description of the Drawings Figure 1a is a topographical view of a memory cell according to the invention showing four possible gain states. Figure 1b is a topographical view of the memory cell of Figure 1a in a high gain state.

Figure 1c is a topographical view of the memory cell of Figure 1a in a medium-high gain state.

Figure 1d is a topographical view of the memory cell of Figure 1a in a medium-low gain state. Figure 1e is a topographical view of the memory cell of Figure 1a in a low-gain state.

Figure 2 is a topographical view of a memory cell having an accelerated taper according to the invention and showing four possible gain states.

Figure 3 is a topographical view of a portion of an array of memory cells according to a preferred embodiment of the invention. Detailed Description of the Drawings Shown in Fig. 1a is a transistor memory cell 10 having four possible gain states, the selection of which is achieved by choosing one of four possible gate areas. The four gate areas are between an upper boundary 12 and one of a high gain line 14, a medium-high gain line 16, a medium-low gain line 18, and a low-gain line 20. The four gain states, in order of decreasing gain, comprise a high gain state, a medium-high gain state, a medium-low gain state, and a low gain state which correspond to gate areas bounded by high gain line 14, medium-high gain line 16, medium-low gain line 18, and low gain line 20, respectively. Each gate area is also between a left boundary 22 and a right boundary 24. A total area available for making a gate is potential gate area 26 which is bounded by upper boundary 12, low gain line 20, left boundary 22, and right boundary 24. In this case potential gate area 26 is the same as the gate area for the low gain state because upper boundary 12 is a boundary for the gate area for each gain state. Left boundary 22 and right boundary 24 as shown are linearly tapered at an angle A from a drain region 28 to a source region 30. Regions 28 and 30 are designated drain and source respectively, for the likely current electrode function of the two regions but can in fact be interchanged so that either region can function as a source or drain. Fig. 1b shows cell 10 in the high gain state with a gate area bounded by left boundary 22, right boundary 24, high gain line 14, and upper boundary 12 over which is applied polysilicon to form a gate. Dopant is then applied in a conventional self-aligning process forming a source and a drain in source and drain regions 30 and 28, respectively. Only a single mask for applying the polysilicon gate is required. Gain of a MOS transistor is directly related to the width to length ratio of the gate. Due to the tapering of the potential gate area toward source region 30, the high gain state cell shown in Fig. 1b does not have a uniform width. Instead, it has a long width W_u at upper boundary 12 and a short width W₁ at high gain line 14. Experiment has shown that an effective width of a tapered gate is a function of the short and long widths, but is more heavily weighted toward the short width. For a linear taper as shown in Fig. 1a, an effective width W_1E is located approximately three-fourths of the way from upper boundary 12 to high gain line 14 and the calculation for obtaining an approximation of effective gate width W_1E is as follows:

A gate length L₁ is measured from upper boundary 12 to high gain line 14. A gain G₁ of the high gain state is directly proportional to -

Fig. 1c shows cell 10 in a medium-high gain state with a gate area bounded by upper boundary 12, medium-high gain line 16, right boundary 24, and left boundary 22 over which is applied polysilicon to form a gate. Dopant is then applied in a conventional self-aligning process forming a source and a drain in source and drain regions

30 and 28, respectively. As in the high gain state depicted in Fig. 1b, only a single mask is required for applying the polysilicon gate. Because medium-high gain line 16 is further along the taper, the gate has a length

L₂ which is longer than length L₁. An effective width

W_1E is related to long width W_u and a short width

W₂ by the following approximation:

A gain G₂ of the medium-high gain state is directly proportional to Gain G₂ is smaller

than Gain G₁ because effective gate width W_2E is less than effective gate width W_1E and gate length L₂ is greater than gate length L ₁. So the gain differential between gains G₁ and G₂ is impacted not by just a difference in one of the gate length and width but by both a decrease in width and an increase in length to obtain the gain decrease.

Figure 1d shows cell 10 in a medium-low gain state with a gate area bounded by upper boundary 12, medium-low gain line 18, right boundary 24, and left boundary 22 over which is applied polysilicon to form a gate. Dopant is then applied in a conventional self-aligning process forming a source and a drain in source and drain regions 30 and 28, respectively. As in the high and medium-high gain states depicted in Figs. 1b and 1c, respectively, only a single mask is required for applying the polysilicon gate. Because medium-low line 18 is further along the taper than medium-high line 16, the gate has a length L₃ which is longer than length L₁. An effective width W_3E is related to long width W_u and a short width W₃ by the following approximation:

A gain G₃ of the medium-low gain state is directly proportional to Gain G₃ is smaller than gain

G₂ because effective gate width W_3E is less than effective gate width W_2E and gate length L₃ is greater than gate length L₂. So the gain differential between gains G₃ and G₂ is impacted not by just a difference in one of the gate length and width but by both a decrease in width and an increase in length to obtain the gain decrease.

Figure 1e shows cell 10 in a low gain state with a gate area bounded by upper boundary 12, low gain line 20, right boundary 24, and left boundary 22 over which is applied polysilicon to form a gate. Dopant is then applied in a conventional self-aligning process forming a source and a drain in source and drain regions 30 and 28, respectively. As in the high, medium-high, and medium-low gain states depicted in Figs. 1b, 1c, and 1d, respectively, only a single mask is required for applying the polysilicon gate. Because low gain line 20 is further along the taper than the medium-low line 18, the gate has a length L₄ which is longer than length L₃. An effective width W_4E is related to long width W_u and a short width W₄ by the following approximation:

A gain G₄ of the low gain state is directly proportional to Gain G₄ is smaller than gain G₃ because

effective gate width W_4E is less than effective gate width W_3E and gate length L₄ is greater than gate length L₃. So the gain differential between gains G₄ and G₃ is impacted by both a decrease in width and an increase in length to obtain the gain decrease. Consequently, with a single program mask, any one of four unique gate areas are formed over potential gate area 26, and each unique gate area has an effective width and length which differ from the other gate areas.

Shown in Fig. 2 is a cell 32 which is substantially the same as cell 10 in Fig. 1a having a source region 34, a drain region 36, an upper boundary 38, a left boundary 40, a right boundary 42, and four gain lines 44, 46, 48, and 50 each for establishing a different gain state. The difference in cell 32 from cell 10 lies in a right boundary 42 which has a discontinuity at point 52 whereat the rate of taper of left and right boundaries 40 and 42 is accelerated. Left boundary 40 could be changed in shape as well but still taper with right boundary 42. Other possible shapes which could taper include but are not limited to curves and staircase shapes. Still other shapes may be used which allow for simultaneous variations in length and width of gate area may be selected, for example a hexagonal shape or a shape defined by a discontinuous curve or line.

Shown in Fig. 3 is a portion of a ROM cell array 310 which uses tapered potential gate areas to form transistor cells of four different gain states. Array 310 has four source regions 312, 314, 316, and 318 of N+ material with metal contacts 320, 322, 324, and 326, respectively, and four drain regions 328, 330, 332, and 334 of N+ material with metal contacts 336, 338, 340, and 342, respectively. Metal contacts 336 and 340 are for connecting drain regions 328 and 332 to a first metal bit sense line. Metal contacts 338 and 342 are for connecting drain regions 330 and 334 to a second metal bit sense line. Metal contacts 320 and 324 are for connecting source regions 312 and 316 to a first metal virtual ground line. Metal contacts 322 and 326 are for connecting source regions 314 and 318 to a second metal virtual ground line.

To the left and right of each source region 312, 314, 316, and 318 are substantially diamond shaped regions of field oxide. Shown are oxide regions 344, 345, 346, 347, 348, 349, 350, 351 and 352. To the left and right of each drain region 328, 330, 332, and 334 are substantially bowtie shaped regions of field oxide. Shown also are oxide regions 354, 355, 356, and 357. The location of oxide regions 344-352 and oxide regions 354-357 establish distinct paths between source regions 312, 314, 316, and 318 and drain regions 328, 330, 332, and 334 in a conventional X cell configuration. The shape, however, of oxide regions 344-352 and oxide regions 354-357 provide a unique taper which allows for formation of one of a high gain, a medium-high gain, a medium-low gain, and a low gain transistor cell between each source 312, 314, 316, and 318 and drain 328, 330, 332, and 334. Each path is tapered at an angle B. For example, the path formed by oxide regions 348 and 357 is tapered from drain region 334 to source region 318 at angle B. Transverse to each path between each source 312, 314, 316 and 318 and drain 328, 330, 332, and 334 is a channel region caused by rows of polysilicon in a conventional self-aligning gate process. Shown are 3 rows, 358, 360, and 362 of polysilicon. Shown are four gate areas 364, 365, 366, and 367 formed by row 358; four gate areas 368, 369, 370, and 371 formed by row 360; and four gate areas 372, 373, 374, and 375 formed by row 362. Each gate area 364-375 determines a transistor of one of high, medium-high, medium-low, and low gain.

For example, a high gain transistor is formed by gate area 373 between source region 316, and drain region 334. Other high gain transistors are formed by gate areas 366, 371, and 374 with the same dimensions as gate area 373. Each gate area 364-375 has a boundary adjacent a drain region and a boundary adjacent a source region. For example, gate area 372 has a drain boundary 376 adjacent drain region 332 and between diamond shaped oxide region 347 and bow tie shaped oxide region 356. A portion of drain boundary 376 extends perpendicular to the longitudinal axis of oxide region 356 from oxide region 356 to a point 378 whereat drain boundary 376 changes direction by 45° to reach diamond shaped oxide 347. Gate area 372 has a drain boundary dimension from bow-tie shaped oxide 356 to point 378 of W_L1 and from point 378 to diamond shaped oxide region 347 of W_L2. Each gate area 364-375 has the same drain boundary dimensions of W_L1 and W_L2 the sum of which is a total drain boundary dimension of W₁. Consequently, the source boundary dimension correlates to the relative gain of the transistors formed by gate areas 364-375. For each gate area 364-375, the source boundary is parallel to the drain boundary.

A source boundary 380 of gate area 373 adjacent to source region 316 has a dimension W_A1 from oxide region 356 to a point 382 and a dimension W_A2 from point 382 to oxide 348. Each high gain gate area 366, 371, 373, and 374 have the same boundary dimensions of W_A1 and W_A2 the sum of which is a total source boundary dimension of W_A. A medium-high gain transistor formed by gate area 365 between source region 312 and drain region 330 has a source boundary 384 of dimensions of W_B1 and W_B2 measured in the same way as for high gain gate area 373. Each medium-high gain gate area 365 and 367 have the same source boundary dimensions W_B1 and W_B2 the sum of which is a total source boundary dimension W_B. A medium-low gain transistor formed by gate area 364 between source-region 312 and drain region 328 has a source boundary 386 of dimensions Wei and W_C2 measured in the same way as for gate areas 373 and 365. Each medium-low gain gate area 364, 369, and 375 have the same source boundary dimensions W_C1 and W_C2 the sum of which is a total source boundary dimension W_C.

A low gain transistor formed by gate 368 between source region 316 and drain region 328 has a source boundary 388 which is measured slightly differently than for other gain states. A portion 390 of gate area 368 is between oxide 354 and source region 316 which has a neglible effect as a gate. Accordingly, source boundary 388 has an effective dimension of W_D between oxide 347 and 354 at their closest points.

Associated with each gain state is a source boundary dimension and a drain boundary dimension with the drain boundary dimension W_L being the same for all the gain states as well as being greater than each of the four source boundary dimensions W_A, W_B, W_C, and W_D. Due to the taper of the potential gate area W_A is greater than W_B which is greater than W_C which is greater than W_D, partially accounting for the reduced gain. Each gain state has an effective gate width proportional to its associated source boundary dimension and the drain boundary dimension. An approximate effective gate width W_AE for the high gain state is calculated as follows:

An approximate effective gate width W_BE for the medium-high gain state is calculated as follows:

An approximate effective gate width W_CE for the medium-low gain state is calculated as follows:

An approximate ef fective gate width W_DE for the low gain state is calculated as follows :

Each gain state has a different gate length as well as a different gate width. A gate length L_A for the high gain states is shown, for example, for gate area 374 bounded by source region 318 and drain region 334. Similarly, a gate length L_B for the medium-high gain states, a gate length L_C for the medium-low gain states, and a gate length L_D for the low gain states are shown, for example, for gate areas 367, 369, and 370, respectively.

For a calculation of the gains of the gain states for comparative purposes, the effective width is divided by the length. Consequently, for the purpose of comparing the relative gains of the four gain-states, the high gain state is the medium-high gain state is

the medium-low gain state is and the low gain

state is These ratios can be manipulated as

desired by varying the lengths L_A, L_B, L_C, and L_D; width portions W_A1, W_A2, W_B1, W_B2, W_C1, W_C2, W_D1 , W_D2, W_L1, and W_L2 for determining the effective gate widths, and angle B of the taper of the potential gate area. In A ROM in which the high gain transistor is biased with a gate voltage of 3.9 volts and a source at essentially ground to produce a drain voltage at approximately 2.9 volts, for sense amplifier detection by a differential amplifier, the following are the dimensions (in microns), for example, for useful gain separation of the four gain states: L_A = 2.0, L_B = 2.5, L_C = 3.4, L_D = 5.25 W_A1 = 3.10, W_A2 = 3.75, W_B1 = 2.75, W_B2 = 3.5 W_C1 = 2.35, W_C2 = 2.9, W_D = 3.25, W_L1 = 4.2, W_L2 = 5.0

The angle B of taper is, for example, 64°.

While the invention has been described in a preferred embodiment, it will be apparent to those skilled in the art that the disclosed invention may be modified in numerous ways and may assume many embodiments other than that specifically set out and described above. Accordingly, it is intended by the appended claims to cover all modifications of the invention which fall within the true spirit and scope of the invention.

Claims

1. A read only memory cell comprising a transistor having a first current electrode, a second current electrode, and a control electrode of a predetermined width and length, wherein the cell is capable of being programmed into more than two states as determined by the width and length of the control electrode, the control electrode having a potential area between the first and second current electrodes, the potential control electrode area characterized as allowing variation of both width and length with a single program mask.

2. The read only memory of claim 1 wherein said potential area between first and second current electrodes is tapered between the first and second current electrodes.

3. The read only memory of claim 1 wherein said potential area between the first and second current electrodes is tapered toward the first current electrodes.

4. A method of programming a read only memory transistor cell, comprising: providing a source region for the transistor; providing a drain region spaced from the source region in a manner to provide a gate region which is tapered between the source and drain regions; and applying a single layer of polysilicon to the gate region to form a gate electrode of a predetermined width and length.

5. An array of transistor memory cells wherein each transistor is capable of being programmed into one of at least three gain states as determined by a gate area thereof of a predetermined effective width and length, comprising: a first transistor programmed to a relatively high gain state, a second transistor programmed to a relatively low gain state, and third transistor programmed to a gain state intermediate that of the gain states of first and second transistors, wherein the first transistor has a gate with an effective width greater and a length less than those of the gate of the third transistor and the third transistor has a gate with an effective width greater and length less than those of the gate of the second transistor.