US 20080048327 A1
Circuitry includes first and second circuits spaced apart by an interconnect region. The interconnect region includes a first interconnect and the second circuit includes a stack of semiconductor layers. The first interconnect extends between the first and second circuits to provide communication therebetween. The second circuit operates as a memory circuit.
1. Circuitry comprising:
a first circuit;
an interconnect region;
a first interconnect included in the interconnect region; and
a second circuit spaced apart from the first circuit by the interconnect region, the second circuit including a stack of semiconductor layers; and
wherein the first interconnect extends between the first and second circuits to provide communication therebetween.
2. The circuitry of
3. The circuitry of
4. The circuitry of
5. The circuitry of
6. The circuitry of
7. Circuitry comprising:
first and second circuits, the second circuit including a vertically oriented semiconductor device which includes a stack of semiconductor layers; and
an interconnect region which includes a first interconnect, the interconnect region extending between the first and second circuits;
wherein the first and second circuits are in communication with each other through the first interconnect.
8. The circuitry of
9. The circuitry of
10. The circuitry of
11. The circuitry of
12. The circuitry of
13. The circuitry of
14. The circuitry of
15. Circuitry comprising:
an interconnect region which includes a first interconnect; and
first and second circuits in communication with each other through the first interconnect, the second circuit including a stack of semiconductor layers;
wherein the second circuit is bonded to the interconnect region.
16. The circuitry of
17. The circuitry of
18. The circuitry of
19. The circuitry of
20. The circuitry of
This application is a divisional of application Ser. No. 11/092,521, entitled “ELECTRONIC CIRCUIT WITH EMBEDDED MEMORY”, filed on Mar. 29, 2005, which claims priority to U.S. Pat. No. 7,052,941 filed on Jun. 21, 2004, the contents of both of which are incorporated herein by reference.
1. Field of the Invention
The present invention relates generally to semiconductor circuitry and, more particularly, to circuitry which includes memory devices.
2. Description of the Related Art
Advances in semiconductor manufacturing technology have provided computer chips with integrated circuits that include many millions of active and passive electronic devices, along with the interconnects to provide the desired circuit connections. As is well-known, most integrated circuits include laterally oriented active and passive electronic devices that are carried on a single major surface of a substrate. Active devices typically include transistors and passive devices typically include resistors, capacitors, and inductors. However, these laterally oriented devices consume significant amounts of chip area.
For example, a typical computer system includes a main computer chip with a processor circuit, a control circuit, and a memory cache that are carried on a single major surface of a substrate. The typical computer system also includes main memory which is positioned on a separate memory chip outside the main computer chip. Since the memory cache is positioned on the same substrate as the processor and control circuits in the main computer chip, it is often referred to as embedded memory.
The memory cache typically includes fast and expensive memory cells, such as Static Random Access Memory (SRAM) cells, and the main memory typically includes slower and less expensive Dynamic Random Access Memory (DRAM) cells. Both SRAM and DRAM cells are larger than the devices included in the processor and control circuits, with SRAM cells being much larger than DRAM cells. As is well-known in the art, cache memory (L1 cache or L2 cache, for example) is used to store information from a slower storage medium or subsystem, such as the main memory or peripherals like hard disks and CD-ROMs, that is accessed frequently to increase the operation of the main computer chip.
One reason the operation of the main computer chip is increased because its idle time is reduced. For example, when the processor circuit accesses the main memory, it does so in about 60 nanoseconds (ns). However, a typical processor circuit can have cycle times of about 2 nanoseconds. Hence, there are about 30 wasted cycles while the processor circuit accesses the main memory. As a result, the processor circuit is idle for many cycle times while it accesses the main memory.
The processor circuit, however, can access the cache memory in about 10 ns to 30 ns, so the idle time is significantly reduced if the information needed is stored in the cache memory. The access time of the processor circuit to a hard disk is even slower at about 10 milliseconds (ms) to 12 ms, and the access time to a CD-ROM drive is about 10 times greater than this. Hence, cache memory uses a small amount of fast and expensive memory to allow the processor circuit faster access to information normally stored by a large amount of slower, less-expensive memory.
With this in mind, it seems like the operation of the computer system can be speeded up even more by increasing the size of the cache memory so that it operates as the main memory or by embedding the main memory on the same substrate as the processor and control circuit and eliminating the cache memory altogether. However, there are several problems with doing this.
One problem with doing this is cost. As mentioned above, the SRAM cells included in cache memory are larger and expensive, so increasing the size of the cache memory would significantly increase the cost of the computer chip. DRAM cells are less expensive and smaller, but to embed them with the main computer chip will still significantly increase the cost. One reason the cost increases for both embedded SRAM and DRAM cells is because the number of masks needed to fabricate the main computer chip increases. For example, to embed SRAM and DRAM memory cells with the main computer chip would require about 3-4 and 6-8 extra masks, respectively. This is because the masks used to fabricate the processor and control circuitry are not compatible with the masks used to fabricate SRAM and DRAM memory cells. Another reason the cost increases is because, as discussed below, the yield in manufacturing computer chips decreases as the size of the computer chip increases.
Another problem is that in today's computer systems, the size of the main memory is much larger than the size of the cache memory. For example, in current state of the art systems, the main memory can store 256 MB to 1 GB in a single memory chip, but the cache memory can only store about 1 MB to 2 MB. This is because the size of the memory circuitry needed to store information in SRAM is much larger than that needed for DRAM. A conventional SRAM circuit includes six transistors to store one bit of information and a conventional DRAM circuit includes one transistor and one capacitor, which tend to be large, to store one bit of information.
For example, the size of a conventional embedded SRAN cell is about 70-120 F2 and the size of a conventional DRAM memory cell is about 15 F2. As is known in the art, 1 F is the minimum photolithographic feature size. Hence, if the computer chip is being fabricated using 90 nm lithography, then 1 F corresponds to 90 nm and 1 F2 corresponds to an area that it 90 nm by 90 nm in size. If the computer chip is being fabricated using 60 nm lithography, then 1 F corresponds to 60 nm and 1 F2 corresponds to an area that it 60 nm by 60 nm in size. Thus, to increase the size of the cache memory by increasing the number of SRAM cells included therein would significantly increase the size of the computer chip and decrease its yield. Further, most of the area on the computer chip will be occupied by memory circuitry instead of processor and control circuitry.
This presents several problems. As mentioned above, one problem is that the yield of computer chips in a manufacturing run decreases as their size increases. As is well-known in the art, several computer chips are fabricated from a single large wafer in a run. The individual computer chips carried by the wafer are typically referred to as die. Once the computer chips are fabricated, the die in the wafer are diced to provide separate chips. A wafer, however, has defects distributed throughout it surface which can negatively impact the operation of the computer chips. If the computer chip is larger in size, then it is more likely to include a defect from the wafer and if the computer chip is smaller in size, then it is less likely to include a defect from the wafer. Hence, smaller computer chips are less likely to be defective. Further, if the computer chip is smaller in size, then more of them can be fabricated from a single wafer, which also decreases costs. Hence, smaller computer chips increase the yield and decrease the costs.
Another problem is that it is typically desirable to increase the number of devices included in the processor and control circuitry so that the processor can operate faster and perform more complicated operations. It is desirable for computer chips to be fast so they can process more data in a given amount of time. The speed of operation of a computer chip is typically measured in the number of instructions per second it can perform.
Computer chips can be made to process more data in a given amount of time in several ways. In one way, the computer chip can include devices which are smaller, but this requires advances in lithography and increasingly expensive manufacturing equipment. As discussed above, they can also be made faster by decreasing the time it takes to perform certain tasks, such as storing or retrieving information to and from memory or other peripherals and subsystems.
Computer chips can also be made faster by increasing the number of devices included therein so that more information can be processed in a given period of time. For example, if one processor operates on 32-bit data, then another processor that operates on 64-bit data can process information twice as fast because it can perform more instructions per second. However, the 64-bit processor will need more devices since there are more bits to process at a given time. Hence, if most of the area on the computer chip is occupied by memory cells, then there is less area for the processor and control circuitry to process data with a higher number of bits. The total area of the computer chip can be increased, but as discussed above, this decreases the yield and increases the cost.
Accordingly, it is highly desirable to provide new structures and methods for fabricating computer chips which operate faster and cost effective to fabricate.
The present invention provides circuitry which includes first and second circuits and an interconnect region. A first interconnect is included in the interconnect region. The first interconnect extends above and between the first and second circuits to provide communication therebetween. A third circuit is positioned on the interconnect region and a second interconnect is included in the interconnect region. The second interconnect extends between the third circuit and at least one of the first and second circuits to provide communication therebetween.
The present invention also provides circuitry which includes control and digital circuitry positioned on a substrate. An interconnect region is positioned on surfaces of the control and digital circuitry to provide communication therebetween. Memory circuitry is positioned on the interconnect region so that the interconnect region provides communication between the memory circuitry and the control circuitry.
The present invention further provides circuitry which includes a substrate and first and second processor circuits carried by the substrate. A control circuit is carried by the substrate and positioned near the first and second processor circuits. An interconnect region is carried by and extends from the substrate. The interconnect region allows the first and second processor circuits to communicate with each other and the control circuit. A memory circuit is positioned on the interconnect region so that the memory circuit is electrically coupled to the control circuit through the interconnect region.
These and other features, aspects, and advantages of the present invention will become better understood with reference to the following drawings, description, and claims.
In this embodiment, computer chip 100 includes processor and control circuits which are carried by a substrate and coupled together so that signals can flow between them. Computer chip 100 also includes a memory circuit positioned above the processor and control circuits. The memory circuit is spaced apart from the processor and control circuits by an interconnect region. The memory circuit is also coupled to the control circuit through the interconnect region so that signals can flow therebetween. In this way, the control circuit can receive output signals from the processor circuit and, in response, provide signals to and receive signals from the memory circuit. In response, the control circuit provides input signals to the processor circuit. It should be noted that a memory circuit is discussed here and throughout the disclosure for illustrative purposes and that, in other embodiments, the memory circuit can be replaced with other circuitry which can be fabricated in the same or a similar manner.
A processor circuit typically executes a series of machine instructions to process data. It usually includes an ALU (Arithmetic/Logic Unit) to perform mathematical operations like addition, subtraction, multiplication and division. Modern processor circuits typically include floating point processors that can perform extremely sophisticated operations on large floating point numbers. A processor circuit provide commands to the control circuit to move data from one memory location to another in the memory circuit. A processor circuit can also make decisions and jump to a new set of instructions based on those decisions.
One advantage of computer chip 100 is that the memory circuit is positioned above the control and processor circuits which is desirable since the memory circuit typically occupies much more area than the control and processor circuits. In some examples of a typical computer chip where the processor, control, and memory circuits are positioned on the same substrate, the memory circuit can occupy 50% or more of the total area of the chip. An example of this is shown in
It should also be noted that in some embodiments, the processor in the computer chip discussed herein can address memory devices on a separate chip positioned outside the computer chip. Further, in some embodiments, the computer chip can include embedded memory cells on the same surface as the control and processor circuits, in addition to the memory devices positioned above them. These memory devices can include a cache memory and/or ROM devices. For example the ROM devices can operate as a BIOS (Basic Input/Output System) for the computer system.
Another advantage of computer chip 100 is that the memory circuit is positioned closer to the control and processor circuits so that signals can flow therebetween in less time. This speed up operation of computer chip 100 because the access time is reduced and computer chip 100 is idle for fewer cycle times. Still another advantage of circuit 100 is that the control and processor circuits are fabricated with a different mask set than the memory circuit. Hence, the masks are less complicated and less expensive to make. A further advantage is that the memory devices are fabricated from blanket semiconductor layers after they have been bonded to the interconnect region. Hence, the memory devices do not need to be aligned with the processor and/or control circuitry, which is a complicated and expensive process.
In this embodiment, computer chip 100 includes control circuit 143 and processor circuit 144 carried by substrate 142 (
In this embodiment, substrate 142 includes silicon, although it can include other materials which can support the subsequent structures positioned thereon. Other suitable substrate materials include gallium arsenide, indium phosphide, and silicon carbide, among others. It should be noted that substrate 142 can have portions doped n-type or p-type and some portions of substrate 142 can even be undoped. The preferred material for substrate 142 in this invention is single crystalline material which can have defects, but is generally better material quality compared to amorphous or polycrystalline material.
In this example, control circuit 143 and processor circuit 144 include digital circuitry known in the art. However, the digital circuitry is not shown in FIGS. 1 or 2 for simplicity and ease of discussion. The digital circuitry can include electronic devices, such as transistors, which extend into substrate 142 and/or out of substrate 142 through a surface 142 a (
As shown in
In accordance with the invention, the interconnects included in interconnect region 131 can be formed so that signals, such as signal Sa, can flow between the various devices included in processor circuit 144. The interconnects can also be coupled together so that signals can flow between control circuit 143 and processor circuit 144. The interconnects, vias, and contacts can include conductive materials known in the art, such as aluminum (Al), copper (Cu), tungsten (W), titanium (Ti), or a doped semiconductor, among others.
For example, signal Sa can flow between processor circuit 144 and control circuit 143 through an interconnect 138 a included in interconnect region 131, as shown in
In this embodiment, a memory circuit 121 is positioned on interconnect region 131 and bonded to surface 131 a. The bonding can be done in many different ways. For example, the bonding can be done by heating bonding surface 131 a and coupling memory circuit 121 thereto. Since memory circuit 121 is bonded to interconnect region 131 instead of deposited thereon, it can include better quality semiconductor material. One reason the material is better quality is because it is more crystalline. It is more crystalline than polycrystalline material which is typically deposited on dielectric regions when wafer bonding is not used. More information about wafer bonding can be found in co-pending U.S. patent applications titled “SEMICONDUCTOR LAYER STRUCTURE AND METHOD OF MAKING THE SAME,” “SEMICONDUCTOR BONDING AND LAYER TRANSFER METHOD,” and “WAFER BONDING METHOD” filed on an even date herewith by the same inventor and incorporated herein by reference.
Memory circuit 121 includes a bit line 120 a positioned on surface 131 a. A dielectric region 123 is positioned on surface 131 a and bit line 120 a. Bit line vias 124 a extend upwardly therefrom bit line 120 a and through dielectric region 123. The number of bit line vias 124 a depends on the number of devices it is desired to form in memory circuit 121. Each bit line via 124 a is coupled to an electronic device 124.
Electronic device 124 is typically a transistor or a memory device, although it can include other devices. The transistor can be a metal oxide semiconductor field effect transistor (MOSFET) and the memory device can be a negative differential resistance (NDR) static random access memory (SRAM) cell. An NDR SRAM includes a layer structure that operates as a transistor and a layer structure that operates as a thyristor. The transistor and thyristor are coupled together to operate as the NDR SRAM cell. More information regarding the NDR SRAM cell can be found in co-pending U.S. patent application titled “SEMICONDUCTOR MEMORY DEVICE” filed on the same date herewith by the same inventor and incorporated herein by reference.
A reference line via 124 b is coupled to the opposite end of each device 124. Each reference line via 124 b extends upwardly from its corresponding device 124 where it connects to a reference line 120 b. In this way, each device 124 is coupled between bit line and reference lines vias 124 a and 124 b. It should be noted, however, that in other embodiments, line 120 a can be used as a reference line and line 120 b can be used as a bit line. A dielectric region 148 is positioned on dielectric region 123 and reference line 120 b.
Memory circuit 121 and interconnect region 131 include interconnects so that signals can flow between control circuit 143 and bit line 120 a and reference line 120 b. In this particular example, a reference interconnect 136 extends through regions 133 and 123 so that one end is coupled to control circuit 143 and the opposite end is coupled to reference line 120 b. Similarly, a bit interconnect 137 extends through region 133 so that one end is coupled to control circuit 143 and the other end is coupled to bit line 120 a. Hence, control circuit 143 can provide a bit signal to bit line 120 a through interconnect 137 and a reference signal to reference line 120 b through interconnect 136. In this way, control circuit 143 can communicate with the devices included in device structure 124. In other examples, the reference line 120 b can be connected to an outside contact (not shown) which provides a reference voltage or current from outside of the circuit 100.
In operation, various signals, such as signal Sa, can flow between processor circuit 144 and control circuit 143. In response, control circuit 143 provides signals to and receives signals from memory circuit 121 through interconnects 136 and 137. The signals can be to read, write, and/or erase information in memory circuit 121. Control circuit 143 then provides input signals to processor circuit 144. The input signals can be data values stored by memory circuit 121 that processor circuit 144 desires to process.
In this way, the conductivity of regions 125 a, 125 b, and/or 125 c can be adjusted in response to a word signal provided to control terminal 125 e. The word signal is provided by control circuit 143 through a word interconnect 139. Word interconnect 139 is coupled between control circuit 143 and control terminal 125 e and extends through dielectric regions 123 and 133 similar to interconnects 136 and 137. Interconnect 139, insulator region 125 d, and control terminal 125 e are not shown in
Computer chip 102 includes control circuit 143 and processors 144 a-144 d which are carried by substrate 142 (
In chip 101, various signals can flow between processors 144 a-144 d and control circuit 143. For example, signal Sa, Sb, Sc, and Sd can flow between processor circuit 144 a, 144 b, 144 c, and 144 d, respectively, and control circuit 143, as shown in
In a particular example, signal Sa can flow between processor circuit 144 a and control circuit 143 through an interconnect 138 a, as shown in
One advantage of chip 101 is that the distance between control circuit 143 and processors 144 a-144 d is reduced so that they can communicate with each other faster. This increases the speed of computer chip 100. Another advantage is that the design of chip 101 is convenient because each processor circuit 144 a-144 d can have the same or a similar design which simplifies its fabrication.
Here, signals Sa1 and Sb1 flow between control circuit 143 a and processors 144 a and 144 b, respectively. Signals Sa2 and Sc1 flow between control circuit 143 c and processors 144 a and 144 c, respectively. Signals Sb2 and Sd1 flow between control circuit 143 b and processors 144 b and 144 d, respectively. Signals Sc2 and Sd2 flow between control circuit 143 d and processors 144 c and 144 d, respectively.
Signals Sa1, Sb1, Sa2, Sc1, Sb2, Sd1, Sc2, and Sd2 flow between corresponding control circuits and processors through interconnects, similar to interconnects 138 a and 138 b, as described above, in conjunction with
Signals Sa1, Sa2, Sb1, Sb2, Sc1, Sc2, Sd1, and Sd2 can flow between corresponding control circuits 143 a-143 d and corresponding processors 144 a-144 d through interconnects, similar to interconnects 138 a and 138 b, as described above, in conjunction with
The present invention is described above with reference to preferred embodiments. However, those skilled in the art will recognize that changes and modifications may be made in the described embodiments without departing from the nature and scope of the present invention. Various further changes and modifications will readily occur to those skilled in the art. To the extent that such modifications and variations do not depart from the spirit of the invention, they are intended to be included within the scope thereof.
Having fully described the invention in such clear and concise terms as to enable those skilled in the art to understand and practice the same, the invention claimed is: