Integrated scaling and stretching platform for optimizing monolithic and / or heterogeneous integration on single semiconductor die

The monolithic die design with miniaturized transistors and self-aligned connections addresses the area expansion issues in SRAM and logic circuits, allowing for more efficient integration of high-performance circuits and cache levels within the die, overcoming photolithography constraints.

JP2024156822AInactive Publication Date: 2024-11-06INVENTION & COLLABORATION LAB PTE LTD
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
JP2024125159
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2021-11-19
Filing Date
2024-07-31
Publication Date
2024-11-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The scaling of SRAM and logic circuits in monolithic semiconductor dies is hindered by increased area due to interference between metal layouts and latch-up issues, limiting the integration of high-performance circuits like GPUs and FPGAs within the die size constraints of current photolithography technology.

Method used

A monolithic die design utilizing miniaturized transistors with self-aligned connections and insulated n+ and p+ regions, allowing direct connections between gate, source, and drain regions to metal layers without intermediate layers, and placing metal lines below the silicon substrate to reduce interference.

Benefits of technology

This design significantly reduces the area of SRAM and logic circuits, enabling more functional blocks and cache levels on a single die, enhancing processing efficiency and capacity while overcoming photolithography limitations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024156822000001_ABST
    Figure 2024156822000001_ABST
Patent Text Reader

Abstract

To provide an optimized monolithic semiconductor die based on an unified scaling and stretching platform that can effectively reduce the size of SRAM circuits and logic circuits within the monolithic semiconductor die without reducing the minimum feature size.SOLUTION: The die area of a new shrink die 3520 is smaller than the die area of an original SOC (System on Chip) die 3510 having a second circuit fabricated on the basis of a first technology node, the first circuit is the same as the second circuit, and the first circuit is an SRAM circuit, a logic circuit, a combination of SRAM and logic circuit, or a major functional block circuit.SELECTED DRAWING: Figure 35
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to monolithic semiconductor dies, and more particularly to an optimized monolithic semiconductor die based on a unified scaling and stretching platform that can effectively reduce the size of SRAM circuits and logic circuits within the monolithic semiconductor die without reducing the minimum feature size. [Background technology]

[0002] IT systems are evolving rapidly in every business and enterprise, including factories, healthcare, and transportation. Now, system on chip (SOC) or artificial intelligence (AI) is the cornerstone of IT systems, making factories smarter, improving patient outcomes, and making autonomous vehicles safer. Data from manufacturing equipment, sensors, and machine vision systems can easily reach a petabyte per day. So, to handle this petabyte-scale data, high performance computing (HPC), SOC, or AI chips are required.

[0003] In general, AI chips can be classified into Graphic Processing Units (GPUs), Field Programmable Gate Arrays (FPGAs), and Application Specific ICs (ASICs). GPUs were originally designed for parallel graphic processing, but they are increasingly being used for AI learning. The learning speed and efficiency of GPUs is generally said to be 10 to 1000 times faster than general-purpose CPUs. FPGAs have blocks of interacting logic that engineers can design to assist with specific algorithms, making them suitable for AI inference. Although FPGAs have the disadvantages of being large in size, slow in speed, and high in power consumption, they are preferred over ASIC designs due to their quick time to market, low cost, and high flexibility. FPGAs are highly flexible, so any part of the FPGA can be partially programmed according to requirements. FPGA inference speed and efficiency are 10 to 100 times faster than general-purpose CPUs. On the other hand, ASICs are generally more efficient than FPGAs because they directly adjust the circuit. In the case of customized ASICs, the learning / inference speed and efficiency are 10 to 1,000 times faster than general-purpose CPUs. However, unlike FPGAs, which are easily customized to match the evolution of AI algorithms, ASICs tend to gradually become obsolete as new AI algorithms are developed.

[0004] Whether it is a GPU, FPGA, or ASIC (or similar SOC, CPU, NPU, etc.), logic and SRAM circuits are the major circuits, and their combination accounts for about 90% of the AI ​​chip size. The remaining 10% of the AI ​​chip may include input / output pad circuits. Nevertheless, scaling process / technology nodes for manufacturing AI chips is becoming increasingly necessary to train AI machines efficiently and quickly, as it provides better efficiency and performance. The improvement of performance and cost of integrated circuits has been mainly achieved by process scaling techniques following Moore's Law, but scaling with technology nodes ("lambda" or "F") and minimum feature sizes from 28nm to 3-5nm has encountered many technical difficulties, and the semiconductor industry's R&D and equipment investment costs have increased dramatically.

[0005] For example, it is becoming increasingly difficult to achieve the scaling of SRAM devices to improve recording density, the reduction of operating voltage (VDD) to reduce standby power consumption, and the improvement of yield, which are necessary to realize large-capacity SRAM. When miniaturized to 28 nm (or less), the manufacturing process becomes an issue. Figure 1A shows the architecture of an SRAM cell, a 6-transistor (6-T) SRAM cell. Two cross-coupled inverters It consists of two access transistors (PMOS pull-up transistors PU-1, PU-2, NMOS pull-down transistors PD-1, PD-2) and two access transistors (NMOS pass-gate transistors PG-1 and PG-2). A high-level voltage VDD is coupled to the PMOS pull-up transistors PU-1, PU-2, and a low-level voltage VSS is coupled to the NMOS pull-down transistors PD-1, PD-2. When a word line (WL) is enabled (i.e., a row is selected in the array), the access transistors turn on and connect the storage nodes (node ​​1 / node 2) to the vertically running bit lines (BL, BL bar). Figure 1B is a "bar graph" that shows the layout and connections of the six transistors in an SRAM. The bar graph usually includes only the active areas (vertical gray bars) and gate lines (horizontal white bars). Of course, there are many more contacts, some of which are directly connected to the six transistors, while others are connected to the word lines (WL), bit lines (BL and BL bar), high-level voltage VDD, low-level voltage VSS, etc.

[0006] As the minimum feature size decreases, λ 2 or F 2Some of the reasons for the dramatic increase in the total area of ​​SRAM cells represented by can be explained as follows. A conventional 6T-SRAM connects six transistors with multiple wirings, the first wiring layer M1 of which connects the gate level (hereafter referred to as "gate") of the transistor with the diffusion level (commonly referred to as "diffusion") of the source and drain regions. Without increasing the die size with only M1, it is necessary to increase the second wiring layer M2 and / or the third wiring layer M3 to facilitate signal transmission (such as word lines (WL) and / or bit lines (BL and BL bar)), and a structure Via1, constructed with a certain seed conductive material, is formed to connect M2 to M1. Thus, there is a vertical structure formed through a contact (Con) connection from the diffusion to M1, i.e. "Diffusion-Contact-M1". Similarly, another structure connecting the gate to M1 with a contact structure can be formed as "Gate-Contact-M1". Also, if it is necessary to form a connection structure such as connecting the M1 wiring to the M2 wiring through Via1, we will call it "M1-Via1-M2". A more complex wiring structure from the gate level to the M2 wiring can be described as "gate-contact-M1-via1-M2". Furthermore, the stacked wiring system can have an "M1-via1-M2-via2-M3" or "M1-via1-M2-via2-M3-via3-M4" structure, etc. Since the gates and diffusions of the two access transistors (NMOS pass gate transistors PG-1 and PG-2 in FIG. 1A) are connected to the word lines (WL) and / or bit lines (BL and BL bar) located in the second wiring layer M2 or the third wiring layer M3, in conventional SRAMs, such metal connections must first pass through the wiring layer M1. That is, in state-of-the-art wiring systems for SRAMs, the gate or diffusion may not be able to connect directly to M2 without bypassing the M1 structure. As a result, the space required between one M1 wiring and another M1 wiring increases the die size and in some cases may thwart the intent of efficient channeling using M2 directly, since this wiring connection exceeds the M1 area.Moreover, it is difficult to form a self-aligned structure between the via 1 and the contact, and at the same time, both the via 1 and the contact are connected to their own wiring systems.

[0007] Furthermore, as shown in Figure 2A, in a conventional 6T-SRAM cell, at least one NMOS transistor and one PMOS transistor are arranged in the adjacent regions of a p-type substrate and an n-type well formed adjacent to each other, so that a junction structure called an n+ / p / n / p+ parasitic bipolar device is formed in the curved line from the n+ region of the NMOS transistor to the p-type well, the adjacent n-type well, and even the p+ region of the PMOS transistor. If a large noise occurs at the n+ / p junction or the p+ / n junction, an abnormally large current may flow through this n+ / p / n / p+ junction, which may cause the operation of a part of the CMOS circuit to stop or the entire chip to malfunction. This abnormal phenomenon, called latch-up, has a negative effect on the operation of CMOS and must be avoided. One way to increase the resistance to latch-up, which is nothing but a weakness of CMOS, is to increase the distance from the n+ region to the p+ region. Therefore, if the distance from the n+ region to the p+ region is increased to avoid the latch-up problem, the size of the SRAM cell will also increase.

[0008] Even if the manufacturing process is scaled down to 28 nm or less (the so-called "minimum feature size", "λ", "F"), due to the above problems such as the size of the contacts, the interference between the layout of metal lines connecting the word lines (WL), bit lines (BL and BL bar), high-level voltage VDD, low-level voltage VSS, etc., as shown in Figure 2B, when the minimum feature size is reduced, λ 2 or F 2 This dramatically increases the total area of ​​the SRAM cell, which is expressed as:

[0009] A similar situation occurs in the scaling of logic circuits. Scaling logic circuits to increase packing density, reduce operating voltage (Vdd) to reduce standby power consumption, and increase yield are becoming increasingly difficult to achieve, all of which are necessary to realize large capacity logic circuits. Standard cells are basic elements commonly used in logic circuits. Standard cells may have basic logic function cells (inverter cells, NOR cells, NAND cells, etc.). Similarly, even with the scaling of manufacturing processes to 28 nm and below, the interference between the size of contacts and the layout of metal lines causes the λ 2 or F 2 The total area of ​​the standard cell, represented by the square root of ...

[0010] The area size of the inverter cell is equal to X×Y, where X=2×Cpp, Y=Cell_Height, and Cpp is the distance of the Contact Poly Pitch. It can be seen that some active areas or fins (shown as light grey bars and called “dummy fins”) are not utilized for the PMOS / NMOS of this standard cell, the potential reason for which may be related to the latch-up issue between PMOS and NMOS. Thus, the latch-up distance between PMOS and NMOS in Figure 3A is 3×Fp, where Fp is the fin pitch. Based on the available data on Cpp (54nm) and Cell_Height (216nm) of the Samsung 5nm (UHD) standard cell, the cell area is X×Y = 23328nm 2 (or 933.12λ 2 , lambda (λ) can be calculated to be equal to the smallest feature size such as 5nm. Figure 3B shows a Samsung 5nm (UHD) standard cell and its dimensions. As shown in Figure 3B, the latch-up distance between PMOS and NMOS is 15λ, Cpp is 10.8λ, and the cell height is 43.2λ.

[0011] Figure 3C shows the scaling trend in area size (2Cpp × cell height) by process technology node for three semiconductor foundries. As the technology node gets smaller (e.g., from 22 nm to 5 nm), λ 2 It is clear that the size of the area of ​​a conventional standard cell (2Cpp × Cell_Height) increases dramatically. For conventional standard cells, the smaller the technology node, the smaller the area of ​​the λ 2 The area size represented by λ becomes larger. 2 The reason why increases sharply is thought to be that, in both SRAM and logic circuits, when λ becomes smaller, the sizes of the gate contact / source contact / drain contact are difficult to reduce proportionally, the latch-up distance between PMOS and NMOS is difficult to reduce proportionally, and metal layers interfere with each other when λ becomes smaller.

[0012] From another perspective, today, SOC, AI, Network Processing Unit (NPU), GPU, CPU, FPGA, etc. are integrated monolithically to accommodate as many circuits as possible. However, as shown in Figure 4A, maximizing the die area of ​​each monolithic die is limited by the maximum reticle size of the lithography stepper, which is difficult to scale due to the state-of-the-art existing photolithography exposure tools. For example, as shown in Figure 4B, the current i193 and EUV lithography steppers have a maximum reticle size, so the scanner maximum field area (SMFA) of a monolithic SOC die is 26 mm × 33 mm or 858 mm. 2 (https: / / en.wikichip.org / wiki / mask). However, for AI applications, high-end consumer GPUs are 500-600mm 2 As a result, it is becoming more difficult or impossible to integrate two or more major functional blocks, such as a GPU and an FPGA (for example), on a single monolithic die within the limitations of the SMFA. Also, the SRAM cells in the most widely used 6-transistor CMOS are quite large to make the size of the eSRAM in both major blocks large enough. Furthermore, there is a need to scale up the capacity of external DRAM, but it is becoming increasingly difficult to scale up the size of discrete PoP (package-on-package, e.g. HBM to In SOC (SOC) or POD (Package DRAM on SOC Die), the constraints of poor signal routing between dies and between packages make it difficult to achieve the desired performance. [Prior art documents] [Non-patent literature]

[0013] [Non-Patent Document 1] J. Zhang et al., "15.1 A5nm 135Mb SRAM in EUV and High-Mobility-Channel FinFET Technology with Metal Coupling and Charge-Sharing Write-Assist Circuitry Schemes for High-Density and Low-VMIN Applications," 2020 IEEE International Semiconductor Circuits Conference (ISSCC), 2020, pp. 238-240. Summary of the Invention [Problem to be solved by the invention]

[0014] Therefore, it is necessary to propose an optimal monolithic / heterogeneous integration structure of a single semiconductor die that solves the above problems in order to optimize the dimensions of standard cells / SRAM cells of a monolithic SOC die within the limitations of SMFA and realize more powerful and efficient SOC and AI single chips in the near future without shrinking the technology node or minimum feature size λ. [Means for solving the problem]

[0015] An embodiment of the present invention provides a first monolithic die, the first monolithic die comprising a first circuit fabricated based on a first technology node, a die area of ​​the first monolithic die being smaller than a die area of ​​a second monolithic die having a second circuit fabricated based on the first technology node, the first circuit being the same as the second circuit, and the first circuit being an SRAM circuit, a logic circuit, a combination of SRAM and logic circuit, or a major functional block circuit.

[0016] According to another aspect of the invention, the second circuit occupies between 20% and 90% of the die area of ​​the second monolithic die.

[0017] According to another aspect of the present invention, the first circuit is a Ynm 2 and the second circuit occupies X nm of the second monolithic die. 2 and X>Y.

[0018] According to another embodiment of the invention, Y is between 20% and 90% of X.

[0019] Yet another embodiment of the present invention is to provide a first monolithic die, the first monolithic die comprising: a first circuit formed on the first monolithic die; and a second circuit formed on the first monolithic die, the first monolithic die having a first scanner maximum exposure area, the first circuit occupying a first portion of the first scanner maximum exposure area, the second circuit occupying a second portion of the first scanner maximum exposure area, the scanner maximum exposure area of ​​the first monolithic die being the same as a second scanner maximum exposure area of ​​the second monolithic die, the second monolithic die having the first circuit, and an area of ​​the first circuit on the second monolithic die being between 50% and 90% of the second scanner maximum exposure area of ​​the second monolithic die.

[0020] According to another aspect of the invention, the first circuit is an SRAM circuit, a combination of an SRAM circuit and a logic circuit, or a major functional block circuit.

[0021] According to another aspect of the invention, the first circuit is selected from the group consisting of a GPU, a CPU, a TPU, an NPU, and an FPGA.

[0022] According to another aspect of the present invention, the first scanner maximum exposure area or the second scanner maximum exposure area is 26 mm x 33 mm or 858 mm 2 The following is the result.

[0023] Yet another embodiment of the present invention provides a first monolithic die comprising a first circuit fabricated by performing a predetermined manufacturing process based on a first technology node, the die area of ​​the first monolithic die being smaller than a die area of ​​a second monolithic die comprising a second circuit fabricated based on the first technology node without performing the predetermined manufacturing process.

[0024] According to another aspect of the invention, the first circuit is the same as the second circuit.

[0025] According to another aspect of the invention, the second circuit occupies between 20% and 90% of the die area of ​​the second monolithic die.

[0026] According to another aspect of the present invention, the first circuit is a Ynm 2 and the second circuit occupies X nm of the second monolithic die. 2 and X>Y.

[0027] According to another embodiment of the present invention, Y is 20% to 90% of X.

[0028] According to another aspect of the invention, the first circuit is an SRAM circuit, a logic circuit, a combination of SRAM and logic circuit, or a major functional block circuit.

[0029] Yet another embodiment of the present invention is to provide a monolithic die comprising: a first processing unit circuit formed on the monolithic die, the first processing unit circuit comprising a plurality of first logic core circuits, each first logic core circuit corresponding to a first cache memory; and a second processing unit circuit formed on the monolithic die, the first processing unit circuit comprising a plurality of second logic core circuits, each second logic core circuit corresponding to a second cache memory, wherein a scanner maximum exposure region area of ​​the monolithic die is defined by a particular technology node.

[0030] According to another aspect of the invention, the scanner maximum exposure area for a monolithic die is 26 mm x 33 mm or 858 mm 2 The following is the result.

[0031] According to another aspect of the invention, the primary functions performed by the first processor circuit are different from the primary functions performed by the second processor circuit.

[0032] According to another aspect of the present invention, the first processing unit circuit or the second processing unit circuit is selected from the group consisting of a GPU, a CPU, a TPU, an NPU, and an FPGA.

[0033] According to another aspect of the invention, the monolithic die further comprises a third cache memory, the third cache memory being configurable and utilized by the first processing unit circuitry and the second processing unit circuitry during operation of the monolithic die.

[0034] According to another aspect of the present invention, the first cache memory, the second cache memory, and the third cache memory are constructed of SRAM.

[0035] Yet another embodiment of the present invention is to provide a monolithic die comprising: a first processing unit circuit formed on the monolithic die, the first processing unit circuit comprising a plurality of first logic core circuits, each first logic core circuit corresponding to a first cache memory; and a second processing unit circuit formed on the monolithic die, the first processing unit circuit comprising a plurality of second logic core circuits, each second logic core circuit corresponding to a second cache memory, wherein a primary function performed by the first processing unit circuit is different from a primary function performed by the second processing unit circuit.

[0036] According to another aspect of the present invention, the first processing unit circuit or the second processing unit circuit is selected from the group consisting of a GPU, a CPU, a TPU, an NPU, and an FPGA.

[0037] According to another aspect of the invention, the monolithic die further comprises a third cache memory, the third cache memory being shared and utilized by the first processing unit circuitry and the second processing unit circuitry during operation of the monolithic die.

[0038] Yet another embodiment of the present invention provides a monolithic die, the monolithic die comprising a first processing unit circuit formed in the monolithic die, the first processing unit circuit comprising a plurality of first logic core circuits, each first logic core circuit corresponding to a lower level cache memory, and a higher level cache memory utilized by the first processing unit circuit, the sum of the size of all of the lower level cache memories and the size of the higher level cache memory being at least 64MB.

[0039] According to another aspect of the invention, the scanner maximum exposure field area of ​​a monolithic die is defined by the photolithography exposure tool utilized for the technology process node.

[0040] According to another aspect of the invention, the scanner maximum exposure area for a monolithic die is 26 mm x 33 mm or 858 mm 2 The following is the result.

[0041] According to another aspect of the invention, the low level cache memory includes an L1 cache and an L2 cache, and the size of the L2 cache is greater than the size of the L1 cache.

[0042] According to another aspect of the invention, the high level cache memory includes an L3 cache shared and utilized by a plurality of first logic core circuits.

[0043] According to another aspect of the present invention, the monolithic die further comprises a second processing unit circuit formed on the monolithic die, the second processing unit circuit comprising a plurality of second logic core circuits, each second logic core circuit corresponding to a second cache memory, the higher level cache memory including an L4 cache shared and utilized by the first processing unit circuit and the second processing unit circuit.

[0044] According to another aspect of the invention, the L4 cache is shared and utilized by the first and second processing unit circuits depending on the settings of a mode register.

[0045] Yet another embodiment of the present invention provides a method for manufacturing a monolithic die having a set of CMOS circuits, the method including: performing a first set of steps to resize a plurality of transistors in the set of CMOS circuits of the monolithic die based on a conventional technology node; and performing a second set of steps to wire the plurality of transistors to other metal layers in the monolithic die based on the conventional technology node, wherein a new size of the monolithic die manufactured by the first set of steps and the second set of steps is smaller than an original size of another monolithic die having the same set of CMOS circuits manufactured based on the conventional technology node without performing the first set of steps and the second set of steps.

[0046] According to another aspect of the invention, a new size of the monolithic die produced by the first set of steps and the second set of steps is less than 50% of an original size of another monolithic die.

[0047] According to another aspect of the present invention, a new size of a monolithic die manufactured by the first set of processes and the second set of processes is less than 35% of an original size of another monolithic die when the conventional technology node is 5 nm.

[0048] Yet another embodiment of the present invention provides a method for manufacturing a monolithic die having a scanner maximum exposure area area defined by a conventional technology node, the method including: performing a first set of processes to resize a plurality of transistors in the monolithic die based on the conventional technology node; and performing a second set of processes to wire the plurality of transistors to other metal layers in the monolithic die based on the conventional technology node, wherein an SRAM capacity of the monolithic die manufactured by the first set of processes and the second set of processes is greater than a capacity of the monolithic die manufactured based on the conventional technology node without performing the first set of processes and the second set of processes.

[0049] According to another aspect of the present invention, a monolithic die having a scanner maximum exposure area area defined by a conventional technology node has a first processing device circuit and a second processing device circuit, where a primary function performed by the first processing device circuit is different from a primary function performed by the second processing device circuit.

[0050] Yet another embodiment of the present invention provides a platform for reconfiguring the memory architecture of a chip system, the chip system comprising a first monolithic die coupled to a first DRAM memory having a first predetermined capacity, the first monolithic die including a first logic circuit and a first SRAM memory, the platform comprising a second monolithic die coupled to a second DRAM memory having a second predetermined capacity, the second monolithic die including a second logic circuit and a second SRAM memory. The first monolithic die has a scanner maximum exposure area based on a targeted technology node, the second monolithic die has a scanner maximum exposure area based on a targeted technology node, the first logic circuit is the same as the second logic circuit, an area of ​​the first logic circuit on the first monolithic die is greater than an area of ​​the second logic circuit on the second monolithic die, a capacity of the first SRAM memory is less than a capacity of the second SRAM memory, and a second predetermined capacity of the second DRAM memory is less than a first predetermined capacity of the first DRAM memory.

[0051] According to another aspect of the present invention, the second DRAM memory is external to the second monolithic die, and the second DRAM memory and the second monolithic die are encapsulated in a single package. Furthermore, the second DRAM memory is an embedded DRAM chip. Furthermore, the first DRAM memory is external to the first monolithic die, and the first DRAM memory and the first monolithic die are encapsulated in another single package. Furthermore, the first DRAM memory is an embedded DRAM chip.

[0052] According to another aspect of the invention, the sum of the area of ​​the first logic circuit and the area of ​​the first SRAM memory occupies at least 80%-90% of the scanner maximum exposure area of ​​the first monolithic die, and the sum of the area of ​​the second logic circuit and the area of ​​the second SRAM memory occupies at least 80%-90% of the scanner maximum exposure area of ​​the second monolithic die. The second DRAM memory is an embedded DRAM chip external to the second monolithic die, and the second DRAM memory and the second monolithic die are enclosed in a single package.

[0053] The advantages and spirit of the present invention can be understood from the following description taken in conjunction with the accompanying drawings: These and other objects of the present invention will no doubt become obvious to those skilled in the art after reading the following detailed description of the preferred embodiment that is illustrated in the various figures and drawings. [Brief description of the drawings]

[0054] [Figure 1A] This is a circuit diagram of a normal 6T-SRAM. [Figure 1B] 1B is a bar graph corresponding to the 6T-SRAM in FIG. 1A, where the active areas correspond to the vertical bars and the gate lines correspond to the horizontal bars. [Figure 2A] 1 is a cross-sectional view of a conventional NMOS and PMOS structure. [Figure 2B] FIG. 1 shows the total area of ​​an SRAM cell in units of λ2 (or F2) for different process dimensions λ (or F) according to currently available manufacturing processes. [Figure 3A] A bar graph showing the layout and connections of PMOS and NMOS transistors in a standard cell is shown. [Figure 3B] A bar graph showing the placement and connections of PMOS and NMOS transistors in a standard cell, along with dimensional information, is shown. [Figure 3C] Scaling trends in area size by process technology node for three semiconductor foundries. [Figure 4A] FIG. 1 illustrates a monolithic SOC die with a scanner maximum exposure field area (SMFA) limited by the maximum reticle size of a lithography stepper. [Figure 4B] FIG. 1 illustrates a monolithic SOC die with a scanner maximum exposure field area (SMFA) limited by the maximum reticle size of a lithography stepper. [Diagram 5] FIG. 1 illustrates a top view of a miniaturized metal-oxide-semiconductor field effect transistor (mMOSFET) in accordance with the present invention. [Figure 6] FIG. 2 shows a cross-section of a pad oxide layer, a pad nitride layer on a substrate, and an STI (Shallow Trench Isolation) oxide 1 formed on the substrate. [Figure 7] FIG. 2 shows a true gate (TG) and a dummy shield gate (DSG) formed above an active area. [Figure 8] FIG. 1 shows how a spin-on dielectric (SOD) is deposited and a well-designed gate mask layer is deposited and etched. [Figure 9] 1 shows the nitride layer over the dummy shield gate (DSG) with the DSG, a portion of the dielectric insulator corresponding to the DSG, and the p-type substrate 102 corresponding to the DSG removed. [Figure 10] FIG. 2 shows the removal of the gate mask layer, etching of the SOD, and deposition of an oxide 2 layer to form STI oxide 2. [Figure 11] FIG. 1 shows deposition of oxide 3 layer, etching to form oxide 3 spacers, formation of lightly doped drains (LDDs) in p-type substrate, deposition of nitride layer, etching back to form nitride spacers, and removal of dielectric insulator. [Figure 12] FIG. 1 shows an intrinsic silicon electrode being grown by Selective Epitaxy Growth (SEG) technique. [Figure 13] FIG. 13 shows how a CVD-STI oxide 3 layer is deposited and etched back to remove the intrinsic silicon electrodes and form the source (n+ source) and drain (n+ drain) of the mMOSFET. [Figure 14] FIG. 13 shows oxide spacers being deposited and etched to form contact hole openings. [Figure 15A]FIG. 1 shows how an SOD layer is deposited to fill the pores in the substrate and the surface is planarized by chemical mechanical polishing (CMP). [Figure 15B] FIG. 15B is a top view of FIG. 15A. [Figure 16] FIG. 15C shows a photoresistive layer formed over the structure in FIG. 15B. [Figure 17] FIG. 13 illustrates an anisotropic etching technique to remove the nitride cap layer in the exposed gate extension region, revealing the conductive metal gate layer. [Figure 18A] FIG. 13 shows the state after removing the photoresistance layer and the SOD layer to form opening regions over both the source and drain regions and forming spacers. [Figure 18B] FIG. 18B is a top view of FIG. 18A. [Figure 19A] FIG. 2 is a diagram showing how a wiring network layer of metal 1 is formed. [Figure 19B] FIG. 19B is a top view of FIG. 19A, in which the gate is connected to the source region through the metal 1 layer. [Figure 20] 1A-1C are a series of top and cross-sectional views illustrating a method of manufacturing a device having a transistor structure according to an embodiment of the present disclosure. [Figure 21] 1A-1C are a series of top and cross-sectional views illustrating a method of manufacturing a device having a transistor structure according to an embodiment of the present disclosure. [Figure 22] 1A-1C are a series of top and cross-sectional views illustrating a method of manufacturing a device having a transistor structure according to an embodiment of the present disclosure. [Figure 23] 1A-1C are a series of top and cross-sectional views illustrating a method of manufacturing a device having a transistor structure according to an embodiment of the present disclosure. [Figure 24-1] 1A and 1B are top and cross-sectional views illustrating another device having a transistor structure according to yet another embodiment of the present disclosure. [Figure 24-2] 1A and 1B are top and cross-sectional views illustrating another device having a transistor structure according to yet another embodiment of the present disclosure. [Figure 25A]FIG. 2 is a cross-sectional view of a PMOS transistor and an NMOS transistor used in an SRAM cell. [Figure 25B] FIG. 2 is a cross-sectional view of a PMOS transistor and an NMOS transistor used in an SRAM cell. [Figure 26A] FIG. 25C is a top view showing a combined structure of the novel PMOS 52 and the novel NMOS 51 shown in FIGS. 25A and 25B. [Figure 26B] 26B is a diagram showing a cross section of the combination of novel PMOS 52 and novel NMOS 51 along the cutting line (Y-axis) of FIG. 26A. [Figure 27] FIG. 1 illustrates a cross section of one combination of conventional PMOS and NMOS transistors. [Figure 28A] FIG. 25C is a top view showing another combination structure of the novel PMOS 52 and the novel NMOS 51 shown in FIGS. 25A and 25B. [Figure 28B] 28B is a diagram showing a cross section of a combination of a novel PMOS 52 and a novel NMOS 51 along the cutting line (X-axis) in FIG. 28A. [Figure 29] FIG. 2 illustrates a cross section of another combination of conventional PMOS and NMOS transistors. [Diagram 30] FIG. 2 is a top view showing another combined structure of PMOS and NMOS transistors used in the novel SRAM cell. [Figure 31A] FIG. 3 is a diagram showing a "bar graph" such as that shown in FIG. 2. [Figure 31B] 1 is a bar graph of the novel 6T SRAM with dimensions. [Figure 31C] 1 is a bar graph of a novel 6T SRAM with contacts according to the present invention. [Diagram 32] The SRAM cell area (denoted as λ2) from three different semiconductor foundries A, B, and C and across different technology nodes of the present invention are shown. [Figure 33A] 1 is a bar graph of the novel 6T SRAM with dimensions. [Figure 33B]1 is a bar graph of a novel 6T SRAM with contacts according to the present invention. [Figure 34A] 1 shows the area (denoted as λ2) of standard cells from three different semiconductor foundries A, B, and C, and across different technology nodes of the present invention. [Figure 34B] Cpp, Fin Pitch, and Cell_Height are shown from three different semiconductor foundries and across different technology nodes of the present invention. [Diagram 35] FIG. 1 illustrates the integrated scaling and stretching platform proposed by the present invention. [Figure 36A] 1A-1D show different embodiments of a monolithic die based on the proposed integrated scaling and stretching platform. [Figure 36B] 1A-1D show different embodiments of a monolithic die based on the proposed integrated scaling and stretching platform. [Figure 36C] 1A-1D show different embodiments of a monolithic die based on the proposed integrated scaling and stretching platform. [Figure 36D] 1A-1D show different embodiments of a monolithic die based on the proposed integrated scaling and stretching platform. [Figure 36E] 1A-1D show different embodiments of a monolithic die based on the proposed integrated scaling and stretching platform. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0055] As mentioned above, in conventional SRAM or logic cells, the size of the transistors did not decrease proportionately when the minimum feature size or technology node was scaled down to 28nm or below. Hereinafter, "technology node" refers to a specific semiconductor manufacturing process announced by a semiconductor foundry (e.g., N5, N7 announced by Taiwan Semiconductor Manufacturing Company) or related data announced by a third party (e.g., wikichip, https: / / en.wikichip.org / ). Different nodes often have different circuit generations and architectures. In general, the smaller the technology node, the smaller the feature size, resulting in smaller transistors that are faster and more power efficient. The term "minimum feature size" is synonymous with the term "technology node". In the semiconductor industry, the terms "contact poly pitch (or Cpp)" and "fin pitch" are often used. "Fin width" refers to the width of the bottom surface of the fin structure of a FinFet transistor or a tri-gate transistor.

[0056] The present invention discloses a miniaturized transistor structure in which the linear dimensions of the source, drain, and gate of the miniaturized transistor can be precisely controlled and the linear dimensions can be made as small as lambda (λ), which is the minimum feature size. Therefore, when two adjacent transistors are connected through the drain / source, the distance between the edges of the gates of the two adjacent transistors can be as small as 2λ. In addition, the linear dimensions of the contact holes for the source, drain, and gate can be made less than λ, such as 0.6λ to 0.8λ.

[0057] FIG. 5 is an example of a miniaturized metal-oxide-semiconductor field effect transistor (mMOSFET) 100 in accordance with the present invention. As shown in FIG. 5, mMOSFET 100 includes: (1) gate structure 101 having length G(L) and width G(W); (2) on the left side of gate structure 101, source 103 having length S(L) and width S(W), which are linear dimensions from the edge of gate structure 101 to the edge of isolation region 105; (3) on the right side of gate structure 101, drain 107 having length D(L) and width D(W), which are linear dimensions from the edge of gate structure 101 to the edge of isolation region 105; (4) in the center of source 103, contact hole 109 formed by a self-aligned technique has an opening length and width denoted as CS(L) and CS(W), respectively; and (5) in the center of drain 107, contact hole 111 formed by a self-aligned technique has an opening length and width denoted as CD(L) and CD(W), respectively. The length G(L), the length D(L), and the length S(L) can be precisely controlled within the range of the minimum feature size λ. Furthermore, the length and width of the openings denoted as CS(L) and CS(W) or the length and width of the openings denoted as DS(L) and DS(W) may be less than λ, for example, 0.6λ to 0.8λ.

[0058] The following is a brief description of the manufacturing process of the mMOSFET 100. A detailed description of the structure of the mMOSFET 100 and its manufacturing process is provided in U.S. patent application Ser. No. 17 / 138,918, filed Dec. 31, 2020, entitled “Miniaturized Transistors With Controlled Dimensions of Source / Drain and Contact Openings and Related Manufacturing Methods,” the entire contents of which are incorporated herein by reference.

[0059] As shown in Figure 6, a pad oxide layer 302 is formed on the substrate 102, and a pad nitride layer 304 is deposited. Also, the active area of ​​the mMOSFET is defined, and a portion of the silicon material outside the active area is removed to create a trench structure. An oxide 1 layer is deposited in the trench structure and etched back to form a shallow trench isolation (STI oxide 1) 306 below the original horizontal surface of the silicon substrate (hereafter HSS).

[0060] The pad oxide layer 302 and the pad nitride layer 304 are removed to form a dielectric insulator 402 on the HSS. Next, a gate layer 602 and a nitride layer 604 are deposited on the HSS, and the gate layer 602 and the nitride layer 604 are etched to form a dummy shield gate having a desired linear distance to the true gate and true gate (TG) of the mMOSFET, as shown in FIG. 7, the length of the true gate and the dummy shield gate is λ, the length of the dummy shield gate is also λ, and the distance between the edges of the true gate and the dummy shield gate is also λ. Of course, for mitigation purposes, these lengths and distances can be larger than λ depending on requirements.

[0061] Next, deposit spin-on dielectric (SOD) 702 and etch back SOD 702. As shown in Fig. 8, a well-designed gate mask layer 802 is formed by photolithography mask technology. Then, as shown in Fig. 9, anisotropic etching technology is used to remove the nitride layer 604 on the dummy shield gate (DSG), and remove the gate layer of the DSG, a part of the dielectric insulator 402 corresponding to the DSG, and the p-type substrate 102 corresponding to the DSG.

[0062] 10, the gate mask layer 802 is removed, the SOD 702 is etched, and the STI oxide 2 1002 is deposited, followed by etching back. Then, as shown in FIG. 11, the oxide 3 layer is deposited and etched back to form the oxide 3 spacer 1502, the lightly doped drain (LDD) 1504 is formed in the p-type substrate 102, a nitride layer is deposited and etched back to form the nitride spacer 1506, and the dielectric insulator 402 is removed.

[0063] Further, as shown in FIG. 12, a selective epitaxial growth (SEG) technique is used to grow an intrinsic silicon electrode 1602. Then, as shown in FIG. 13, a CVD-STI oxide 3 layer 1702 is deposited and etched back to remove the intrinsic silicon 1602, forming a source region (n+ source) 1704 and a drain region (n+ drain) 1706 of the mMOSFET. The source region (n+ source) 1704 and the drain region (n+ drain) 1706 form the position originally occupied by the dummy shield gate (DSG) between the true gate (TG) and the CVD-STI oxide 3 layer 1702, so that the length and width of the source region (n+ source) 1704 (or the drain region (n+ drain) 1706) are reduced to λ. The opening of the source region (n+ source) 1704 (or the drain region (n+ drain) 1706) can be less than λ, for example 0.8λ. Such openings can be reduced by further forming oxide spacers 1802 as shown in FIG.

[0064] The novel scaled transistor also allows the first metal wiring (M1 layer) to directly connect gate, source, and / or drain regions through self-aligned scaled contacts without the use of conventional contact hole opening masks or metal 0 transition layers for M1 connections. Following FIG. 13, a layer of SOD 1901 is deposited to fill the holes on the substrate, including the openings for the source region (n+ source) 1704 (or drain region (n+ drain) 1706). The surface is then planarized using CMP as shown in FIG. 15A. FIG. 15B is a top view of FIG. 15A, showing multiple fingers in the horizontal direction.

[0065] Furthermore, using a well-designed mask, the photoresistive layer 1902 which becomes some stripe pattern along the X-axis of FIG. 15(b) is carried out at another space of length GROC(L) to expose the area of ​​the gate extension region along the Y-axis of FIG. 15(b), the result is as shown in the top view of FIG. 16. As shown in FIG. 16, the most aggressive design rule of GROC(L)=λ. Then, anisotropic etching technique is used to remove the nitride cap layer in the exposed gate extension region to reveal the conductive metal gate layer (FIG. 17).

[0066] After that, the photoresistive layer 1902 is removed, and the SOD layer 1901 is removed so that their opening regions on top of both the source region 1704 and the drain region 1706 are exposed again. Next, a layer of oxide 1904 with a well-designed thickness is deposited, and then anisotropic etching techniques are used to form spacers on the opening regions of the source region 1704 and the drain region 1706 and on the four side walls of the exposed gate extension region 1903. Thus, naturally occurring contact hole openings are formed in the exposed gate extension region, the source region 1704, and the drain region 1706, respectively. FIG. 18A is a cross-sectional view of such a transistor structure. FIG. 18B is a top view of such a transistor structure of FIG. 18A. The vertical length CRMG(L) of the opening of the exposed gate extension region 1903 is smaller than the length GROC(L), which may be λ.

[0067] Finally, a layer of metal 1 1905 is formed with a well-designed thickness to fill all the aforementioned contact hole openings and to have a smooth plane following the topography of the wafer surface. Then, all the connections between those contact hole openings are respectively created using photolithographic masking techniques to realize the required metal-1 wiring network as shown in Figure 19(a). Figure 19B is a top view of the mMOSFET shown in Figure 19A. Therefore, this metal 1 layer realizes contact filling and plug connections for both gate and source / drain, as well as direct wiring to connect all the transistors. In particular, it eliminates the need for expensive and very tightly controlled traditional contact hole masks and the subsequent very difficult contact hole opening drilling process, which is the most difficult challenge in further scaling the horizontal geometry of billions of transistors. Moreover, it eliminates the need for complex and integrated processing steps (e.g., required for some of the state-of-the-art technologies to create zero metal structures) to insert metal plugs into the contact hole openings and CMP processes to realize metal studs.

[0068] Furthermore, conventional SRAM or standard cells may not allow a gate or diffusion to be directly connected to M2 without bypassing the M1 structure (or M1 to be connected to M3 without bypassing the M2 structure, or M1 to be connected to Mx without bypassing the M2-Mx-1 structure, etc.). The present invention discloses a novel wiring structure that directly connects either the gate region or the diffusion (source / drain) region to the M2 wiring layer in a self-aligned manner without a transition layer M1 through one vertical conductive plug consisting of contact A and via 1A, which are respectively formed in the construction phase of contact and via 1 at another location on the same die. As a result, the space required between one M1 wire and another M1 wire and blocking problems in some wiring connections can be reduced. In the following, a brief description is given of the novel wiring structure in which the gate region and the diffusion (source / drain) region are directly connected to the M2 wiring layer in a self-aligned manner without a transition layer M1.

[0069] 20A-20C show cross-sections and top views up to a construction stage forming a plurality of opening holes over both the gate extension region and the diffusion region of the transistor, in which FIG. 20A shows a top view of the construction stage of the transistor, and FIG. 20B and FIG. 20C show two cross-sections of the construction stage of the transistor along cut lines C1B1 and C1B2 shown in FIG. 20A, respectively. As shown in FIG. 20B and FIG. 20C, a transistor structure 100 is formed and delimited by a shallow trench isolator (STI) 105. The transistor structure 100 has a gate terminal 102, a transistor channel region 103 below the gate terminal 102, and source / drain regions 104. The gate terminal 102 comprises a gate dielectric layer 102a, a gate conductive layer 102b formed on the gate dielectric layer 102a, and a silicon region (or seed region) 102c formed on the gate conductive layer 102b. The silicon region 102c can be composed of polycrystalline or amorphous silicon. The gate terminal 102 further includes a capping layer (e.g., a nitride layer) on the top surface of the silicon region 102c, and at least one spacer (e.g., including a nitride spacer 102s1 and a thermal oxide spacer 102s2) on the sidewalls of the gate dielectric layer 102a, the gate conductive layer 102b, and the silicon region 102c. A first dielectric layer 120 is formed on the semiconductor substrate 101 to at least cover the active region of the transistor structure 100, including the gate terminal 102 and the source / drain regions 104 and the STI 105.

[0070] A plurality of opening holes (such as opening holes 107a and 107b) are formed in the first dielectric layer 120 to expose the top surface portion 11 of the silicon 102c region and the top surface portion 12 of the source / drain region 104. In some embodiments, the opening holes 107a and 107b are formed by partially removing the first dielectric layer 120 by a photolithography process to expose the portions of the silicon region 102c of the source / drain region 104 and the silicon region of the drain terminal. In one example, each of the opening holes 107a and 107b can be sized equal to a minimum feature (e.g., a critical size of the transistor structure 100 of the device 10). Of course, the size of the opening holes 107a and 107b can be larger than the minimum feature. The bottoms of the open holes 107a and 107b (i.e., the exposed top surface portion 11 and the exposed top surface portion 12) are made of either polycrystalline / amorphous silicon or highly conductive highly doped crystalline silicon material, respectively. The exposed silicon region 102c of the gate terminal and the exposed silicon region of the source / drain terminal are seed regions for growing seed-based pillars by selective epitaxial growth (SEG) technique.

[0071] Next, as shown in FIG. 21A to FIG. 21C, a highly doped conductive silicon plug (or conductor pillar) is grown by SEG based on the exposed upper surface portion 11 and the exposed upper surface portion 12 to form a first conductor pillar portion 131a and a third conductor pillar portion 131b. Then, a first dielectric underlayer 140 is formed on the first dielectric layer 120, so that an upper surface 140s of the first dielectric underlayer 140 is substantially flush with the upper surfaces of the first conductor pillar portion 131a and the third conductor pillar portion 131b. The "exposed heads" (or exposed upper surfaces) of the first conductor pillar portion 131a and the third conductor pillar portion 131b can be used as seed portions for a subsequent SEG process. Furthermore, each of the first conductor pillar portion 131a and the third conductor pillar portion 131b has a seed region or seed pillar at its upper portion, which can be used for the next selective epitaxial growth. Then, the second conductor pillar portion 132a is formed on the first conductor pillar portion 131a by the second selective epitaxial growth, and the fourth conductor pillar portion 132b is formed on the third conductor pillar portion 131b. FIG. 21A is a top view showing a structure after the second conductor pillar portion 132a and the fourth conductor pillar portion 132b are formed on the first conductor pillar portion 131a and the third conductor pillar portion 131b according to an embodiment of the present disclosure. FIG. 21B is a cross-sectional view taken along the cutting line C1E1 depicted in FIG. 21A. FIG. 21B is a cross-sectional view taken along the cutting line C1E2 depicted in FIG. 21A.

[0072] Further, as shown in FIGS. 22A-22C, a first conductive layer 150, such as copper (Cu), aluminum (Al), tungsten (W) or other suitable conductive material, can be deposited on the upper surface 140s of the first dielectric underlayer 140. A second dielectric underlayer 160 is then deposited on the first conductive layer 150. The first conductive layer 150 and the second dielectric underlayer 160 are patterned to define an open hollow 109, and the first conductor pillar 130A penetrates the open hollow 109 without contacting the first conductive layer 150 and the second dielectric underlayer 160. FIG. 22A is a top view of the structure after forming the first conductive layer 150 and the second dielectric underlayer 160 on the first dielectric layer 120 according to an embodiment of the present disclosure. FIG. 22B is a cross-sectional view taken along the cut line C1F1 depicted in FIG. 22A. FIG. 22C is a cross-sectional view taken along section line C1F2 depicted in FIG. 22A.

[0073] Further, as shown in FIGS. 23A to 23C, the upper dielectric layer 170 is deposited to cover the second dielectric underlayer 160 and the first dielectric underlayer 140, filling the open hollow portion 109. The upper surface 170s of the upper dielectric layer 170 is located below the upper surface 130t of the first conductor pillar 130A (including the first conductor pillar portion or sub-pillar 131a and the second conductor pillar portion or sub-pillar 132a) and the second conductor pillar 130B (including the third conductor pillar portion or sub-pillar 131b and the fourth conductor pillar portion or sub-pillar 132b). Then, the upper conductive layer 180 is formed on the upper dielectric layer 170, and the first conductor pillar 130A is connected to the upper conductive layer 180 but is disconnected from the first conductive layer 150. In this example, Figure 23A is a top view of the structure after forming a conductive layer 180 over an upper dielectric layer 170 according to one embodiment of the present disclosure. Figure 23A is a top view, and Figure 23B is a cross-sectional view taken along cut line C1H1 depicted in Figure 23A. Figure 23C is a cross-sectional view taken along cut line C1H2 depicted in Figure 23A.

[0074] As described above, each of the exposed silicon regions 102c of the gate terminal and the exposed silicon regions of the source / drain terminal is a seed region for growing a pillar based on a seed region by selective epitaxial growth (SEG). Furthermore, each of the first conductor pillar portion 131a and the third conductor pillar portion 131b also has a seed region or seed pillar at its upper portion, and such seed region or seed pillar can be used for the next selective epitaxial growth. Also, this embodiment can also make it possible to directly connect the M1 wiring (one seed conductive terminal) or conductive layer with the MX wiring layer in a self-aligned manner through one vertical conductive or conductive plug (without connecting to the conductive layers M2, M3, ..., MX-1) if there is a seed portion or seed pillar at the upper portion of the conductive terminal and the conductive pillar portion configured according to the selective epitaxial growth technology. Note that the seed portion or seed pillar is not limited to silicon, and can be any material that can be used as a seed configured for the subsequent selective epitaxial growth.

[0075] The conductor pillar may be a metal conductor pillar or a composite conductor pillar of a metal conductor pillar and a seed portion or seed pillar at its upper portion. As shown in FIG. 24A-FIG. 24C, the highly doped N+ polycrystalline silicon pillars 131a, 132a, 131b, and 132b in FIG. 23A-FIG. 23C can be removed and replaced with a tungsten pillar 330w, a TiN layer 330n, and a highly doped silicon pillar. As shown in FIG. 24B-FIG. 24C, the first conductor pillar includes a metal pillar portion 330A (including a tungsten pillar 330w and a TiN layer 330n) and a highly doped silicon pillar 410a, and the second conductor pillar includes a metal pillar portion 330B (including a tungsten pillar 330w and a TiN layer 330n) and a highly doped silicon pillar 410b. The heavily doped silicon pillars 410a and 410b are seed regions or seed pillars of a conductor pillar configured for subsequent metal connection, as shown in FIG. 24B and FIG. 24C, and a first conductive layer 450 is formed on the first dielectric underlayer 240 and electrically connected to the heavily doped silicon pillars 410a and 410b. Furthermore, the heavily doped silicon pillars 410a and 410b are seed regions or seed pillars of a conductor pillar configured for subsequent SEG process to grow another silicon pillar thereon. In this example, FIG. 24A is a top view, FIG. 24B is a cross-sectional view taken along the cut line C4B1 depicted in FIG. 24A, and FIG. 24C is a cross-sectional view taken along the cut line C4B2 depicted in FIG. 24A. Thus, the conductor pillar may include a tungsten pillar and a first heavily doped silicon pillar, i.e., the conductor pillar has a seed region or seed pillar at its upper portion.

[0076] The heavily doped silicon pillars 410a and 410b are seed regions or seed pillars of the conductor pillars configured to follow the SEG process to grow another silicon pillar on its upper portion, so that the conductor pillars can have seed regions or seed pillars on their upper side, and borderless contact is achieved. As shown in Figures 24D-24F, even if the width of the metal conductive layer (e.g., the first metal underlayer 550a or the second metal underlayer 550b) is the same as the width of the contact plug below (it may be as small as the smallest feature), the metal conductive layer 550a or 550b may not completely cover the contact (as shown in Figures 24E and 24F) due to the misalignment tolerance of the photolithographic masking, but there is no worry that the resistance between the metal conductive layer and the contact will be too high due to lack of contact area. The invention here is to further use SEG to grow some extra heavily doped silicon material (side pillars 520) to attach the vertical walls of the metal conductive layers 550a and 550b. In this example, Fig. 24D is a top view, Fig. 24E is a cross-sectional view taken along section line C51 depicted in Fig. 24D, and Fig. 24F is a cross-sectional view taken along section line C52 depicted in Fig. 24D.

[0077] Further, the present invention discloses a novel CMOS structure in which the n+ and p+ regions of the source and drain regions in the NMOS and PMOS transistors are fully insulated by an insulator, which not only increases the resistance to latch-up problems but also increases the insulation distance to the silicon substrate to separate the junctions of the NMOS and PMOS transistors and reduces the surface distance between the junctions (e.g., 3λ), thereby reducing the size of the SRAM cell or standard cell. In the following, a novel CMOS structure in which the n+ and p+ regions of the source and drain regions in the NMOS and PMOS transistors are fully insulated by an insulator, respectively, is briefly described. A detailed description of the novel combined PMOS and MNOS structure is provided in U.S. Patent Application No. 17 / 318,097, filed May 12, 2021, and entitled "Complementary MOSFET Structures with Local Insulation in Silicon Substrate to Reduce Leakage and Prevent Latch-Up," the entire contents of which are incorporated herein by reference.

[0078] Please refer to FIG. 25A and FIG. 25B. FIG. 25A is a diagram showing a cross section of a PMOS transistor 52, and FIG. 25B is a diagram showing a cross section of an NMOS transistor 51. A gate structure 33 having a gate dielectric layer 331 and a gate conductive layer 332 (such as a gate metal) is formed above a horizontal or original surface of a semiconductor substrate (such as a silicon substrate). A dielectric cap 333 (such as a composite of an oxide layer and a nitride layer) is provided on the gate conductive layer 332. Furthermore, a spacer 34 including a composite of an oxide layer 341 and a nitride layer 342 is used on the sidewall of the gate structure 33. A trench is formed in the silicon substrate, and all or at least a portion of the source region 35 and the drain region 36 are disposed in the corresponding trench. The source (or drain) region in the PMOS transistor 32 may include a P+ region or other suitable doping profile region (such as a gradual or stepwise change between a P- region and a P+ region). Additionally, a local insulator 48 (such as a nitride or other high-k dielectric material) is disposed in one trench and underlies the source region, and another local insulator 48 is disposed in another trench and underlies the drain region. Such a local insulator 48 is below the horizontal silicon surface (HSS) of the silicon substrate and can be referred to as a localized isolation to silicon substrate (LISS) 48. The LISS 48 can be a composite of a thick nitride layer or a dielectric layer. For example, the localized isolation or LISS 48 can comprise a composite local insulator including an oxide layer (referred to as oxide 3V layer 481) covering at least a portion of the sidewall of the trench and another oxide layer (referred to as oxide 3B layer 482) covering at least a portion of the bottom wall of the trench. The oxide 3V layer 481 and the oxide 3B layer 482 can be formed by a thermal oxidation process. The composite local insulator 48 further includes a nitride layer 483 (referred to as nitride 3) disposed on the oxide 3B layer 482 and in contact with the oxide 3V layer 481. It is noted that the nitride layer 483 or nitride 3V may be replaced with any suitable insulating material so long as the oxide 3V layer remains as designed to the maximum extent possible.Additionally, the STI (shallow trench isolation) region in Figures 25A and 25B may comprise a composite STI 49 including an STI-1 layer 491 and an STI-2 layer 492, each of which may be comprised of a thick oxide material by a different process.

[0079] Furthermore, the source (or drain) region in Figures 25A and 25B may comprise a composite source region 55 and / or drain region 56. For example, as shown in Figure 25(a), in a PMOS transistor 52, the composite source region 55 (or drain region 56) comprises at least a lightly doped drain (LDD) 551 in a trench and a heavily P+ doped region 552. In particular, it is noted that the lightly doped drain (LDD) 551 abuts an exposed silicon surface having a uniform (110) crystal orientation. In contrast to the edge of the gate structure, which is labeled in Figure 25A as TEC (the thickness of the etched transistor body well defined to have a sharp edge with an effective channel length), the vertical boundary of the exposed silicon surface has a suitable recessed thickness. The exposed silicon surface is substantially aligned with the gate structure. The exposed silicon surface may also be the terminal surface of the transistor's channel.

[0080] The lightly doped drain (LDD) 551 and the heavily P+ doped region 552 are formed based on selective epitaxial growth (SEG) technique (or other suitable technique which may be atomic layer deposition ALD or selective growth ALD-SALD) to grow silicon from the exposed TEC area to use as crystal seed to form new, well-organized (110) lattice throughout the LISS region where there is no seeding effect on modifying the (110) crystal structure of the newly formed crystal of the composite source region 55 or drain region 56. Such newly formed crystal (including the lightly doped drain (LDD) 551 and the heavily P+ doped region 552) may be named TEC-Si as noted in FIG. 25A. In one embodiment, the TEC is aligned or substantially aligned with the edge of the gate structure 33, the length of the LDD 551 is adjustable, and the sidewall of the LDD 551 opposite to the TEC may be aligned or substantially aligned with the sidewall of the spacer 34. Similarly, the TEC-Si (including LDD and heavily n+ doped regions) of the composite source / drain region of NMOS transistor 51 is shown in FIG. 25B. The composite source (or drain) region may further comprise several tungsten (or other suitable metal material) plugs 553 formed in horizontal connection to the TEC-Si portions to complete the entire source / drain region, as shown in FIGS. 25A and 25B. As shown in FIG. 25A, the active channel current flowing in a future metal interconnect, such as a metal 1 layer, passes through the LDD 551 and the heavily doped conductive regions 552 to the tungsten 553 (or other metal material) which is directly connected to metal 1 by a good metal-metal ohmic contact, which has much lower resistance than a conventional silicon-metal contact.

[0081] One of the combination structures of the novel PMOS 52 and the novel NMOS 51 is shown in FIG. 26A which is a top view, and FIG. 26B is a cross-sectional view of the combination of the novel PMOS 52 and the novel NMOS 51 along the cutting line (Y-axis) in FIG. 26A. As shown in FIG. 26B, since there is a composite local insulator (or LISS 48) between the bottom of the p+ source / drain region of the PMOS and the n-type N-well, there is also another composite local insulator (or LISS 48) between the bottom of the NMOS's N+ source / drain region and the p-type P-well or substrate. In this novel inventive CMOS structure shown in FIG. 26B, it is clearly shown as an advantage that the n+ region and the bottom of the p+ region are completely isolated by the insulator. That is, the possible latch-up path from the bottom of the p+ region of the PMOS to the bottom of the n+ region of the NMOS is completely blocked by the LISS. On the other hand, in a conventional CMOS structure, as shown in FIG. 27, the n+ and p+ regions are not completely isolated by an insulator, so that the possible latch-up paths that exist from the n+ / p junction through the p-well / n-well junction to the n / p+ junction have lengths of circle a, circle b, and circle c ("Circle X" refers to the letter "X" in a circle). (FIG. 27). Therefore, from the viewpoint of device layout, the edge distance (Xn+Xp) ensured between the NMOS and PMOS in FIG. 26B can be made smaller than that in FIG. 27. For example, the ensured edge distance (Xn+Xp) can be about 2 to 5λ, e.g., about 3λ.

[0082] Another combination structure of the novel PMOS 52 and the novel NMOS 51 is shown in FIG. 28A which is a top view, and FIG. 28B is a cross-sectional view of the combination of the novel PMOS 52 and the novel NMOS 51 along the cutting line (X-axis) in FIG. 28A. As shown in FIG. 28B, the path from the n+ / p junction through the p-well (or p-substrate) / n-well junction to the n / p+ junction is significantly longer. The possible latch-up paths from the LDD-n / p junction through the p-well / n-well junction to the n / LDD-p junction are length circle 1, length circle 2 (length of the bottom wall of one LISS region), length circle 3, length circle 4, length circle 5, length circle 6, length circle 7 (length of the bottom wall of another LISS region), and length circle 8 shown in FIG. 28B. On the other hand, in the conventional CMOS structure combining the PMOS and NMOS structures shown in FIG. 29, the possible latch-up path from the n+ / p junction through the p-well / n-well junction to the n / p+ junction only has length circle d, length circle e, length circle f, and length circle g (as shown in FIG. 29). The possible latch-up path in FIG. 28B is longer than that in FIG. 29. Therefore, from the viewpoint of device layout, the edge-to-edge distance (Xn+Xp) ensured between the NMOS and PMOS in FIG. 28B can be smaller than that in FIG. 29. For example, the ensured edge-to-edge distance (Xn+Xp) can be about 2 to 5λ, for example, about 3λ.

[0083] Furthermore, in currently available SRAM cells and standard cells, the metal lines of the high-level voltage VDD and the low-level voltage VSS (or ground) are arranged above the original silicon surface of the silicon substrate, and such distribution may cause interference if there is not enough space between other word lines (WL), bit lines (BL, BL bar), or other connecting metal lines. The present invention discloses a novel SRAM structure that allows the metal lines for the high-level voltage VDD and / or the low-level voltage VSS to be arranged below the original silicon surface of the silicon substrate, so that the size of the SRAM cell can be reduced while avoiding interference between the size of the contacts, the layout of the metal lines connecting the word lines (WL), bit lines (BL and BL bar), the high-level voltage VDD and the low-level voltage VSS, etc. As shown in FIG. 30, in the drain region of the PMOS 52, the metal material 553 such as tungsten is directly coupled to an N-type well electrically coupled to VDD. Meanwhile, in the source region of the NMOS 51, the metal material 553 such as tungsten is directly coupled to a P-type well or a P-type substrate electrically coupled to ground. Thus, openings for source / drain regions that would otherwise be used to electrically couple the source / drain regions to metal layer 2 or metal layer 3 for VDD or ground connections may be omitted in the novel SRAM cells and standard cells. A detailed description of the structure and its manufacturing process is provided in U.S. patent application Ser. No. 16 / 991,044, filed Aug. 12, 2020, and entitled “Transistor Structure and Related Inverter,” the entire contents of which are incorporated herein by reference.

[0084] As discussed above, the new SRAM cells and standard cells have at least the following advantages: (1) The linear dimensions of the source, drain, and gate of a transistor in an SRAM can be precisely controlled, and the linear dimensions can be made small enough to be equivalent to lambda (λ), which is the minimum feature size. Therefore, when two adjacent transistors are connected through the drain / source, the length dimension of the transistor can be as small as 3λ, and the distance between the edges of the gates of two adjacent transistors can be as small as 2λ. Of course, taking into account tolerances, the length dimension of the transistor needs to be 3λ to 6λ or more, and the distance between the edges of the gates of two adjacent transistors needs to be 3λ to 5λ or more. (2) The first metal wiring (M1 layer) is directly connected to the gate, source, and / or drain regions via self-aligned, scaled contacts, without the use of conventional contact hole opening masks or metal 0 transition layers for M1 connections. (3) The gate and / or diffusion (source / drain) regions are directly connected to the M2 wiring layer in a self-aligned manner without connecting the M1 layer, thereby reducing the space required between one M1 line and another and the blocking issues in some wiring connections. Furthermore, a similar structure can be applied when the lower metal layer and the upper metal layer are directly connected by a conductor pillar, but the conductor pillar is not electrically connected to any intermediate metal layer between the lower metal layer and the upper metal layer. (4) The n+ and p+ regions of the source and drain regions in the NMOS and PMOS transistors, respectively, are completely isolated by an insulator, which not only increases the resistance to latch-up problems but also increases the insulation distance to the silicon substrate to separate the junctions of the NMOS and PMOS transistors and reduces the surface distance between the junctions. (e.g., 3λ to 10λ, e.g., 6 or 8λ). (5) Metal lines for the high-level voltage VDD and / or low-level voltage VSS of the SRAM cells and standard cells can be arranged below the original silicon surface of the silicon substrate, so that interference between the size of contacts and the layout of metal lines connecting the word lines (WL), bit lines (BL and BL bar), high-level voltage VDD and low-level voltage VSS, etc. can be avoided even if the size of the SRAM cells and standard cells is reduced. Furthermore, openings for source / drain regions that are originally used for electrical coupling between the source / drain regions and metal layer 2 or metal layer 3 for VDD or ground connection may be omitted in the new SRAM cells and standard cells.

[0085] FIG. 31A is a reproduction of FIG. 1B showing a "bar graph" representing the layout and connections between the six transistors of the SRAM, and FIG. 31B is a bar graph of dimensions according to the advantages of the present invention. FIG. 31B is a bar graph of a novel 6T STAM having dimensions according to the advantages of the present invention. As shown in FIG. 31B, the length dimension of the transistors can be as small as 3λ (enclosed in the dotted rectangle) and the distance between the edges of the gates of two adjacent transistors can be as small as 2λ. Furthermore, the insulation distance to the silicon substrate to separate the junctions of the NMOS and PMOS transistors can be as small as 3λ (enclosed in the dashed rectangle). The insulation distance to the silicon substrate to separate the junctions of the two PMOS transistors can be between 1.5 and 2.5λ, for example, as small as 2λ (enclosed in the dashed rectangle). Furthermore, FIG. 31B shows that Cpp is as small as 3λ and the two fin pitches Fp_1, Fp_2 are as small as 4λ and 3λ, respectively.

[0086] In FIG. 31B, the dimensions of the active areas (vertical lines) can be as small as λ, as can the gate lines (horizontal lines). Furthermore, in FIG. 31B, the upper left transistor corresponding to the PG transistor in FIG. 31A has a horizontal distance of 1.5λ (surrounded by a two-dot dashed rectangle) between the edge of the active area and the boundary of the SRAM cell or bit cell to avoid interference between the two contact holes that will be formed later in the active area and gate area, respectively. The same is true for the transistor in the lower right corner of FIG. 31B, corresponding to another PG transistor in FIG. 31A. Thus, in the case of the bar graph in FIG. 31B, the horizontal length (x direction) of the SRAM cell is 15λ, and the vertical length (y direction) of the SRAM cell or bit cell is 6λ. Thus, the total area of ​​the SRAM cell or bit cell in FIG. 31B is 90λ. 2 and becomes smaller.

[0087] As shown in Fig. 31C, in the proposed SRAM cell, several source / drain contacts (for connecting to the metal 1 layer) can be formed in the active area. The size of the source / drain contacts can be as small as λ x λ, regardless of the technology node or (or minimum feature size). Similarly, several source / drain contacts and gate contacts (for connecting directly to the metal 2 layer without connecting the metal 1 layer as explained earlier) can be formed on the gate or poly line, and the size of the gate contact can be as small as λ x λ as well.

[0088] Figure 32 shows the SRAM cell area (λ) across different technology nodes for three different semiconductor foundries A, B, and C. 2 As feature size shrinks, the SRAM cell size (λ 2 It can be seen that the SRAM cell area across different technology nodes can be kept flat or low sensitivity to technology nodes by the design described in this invention and its derivatives, which means that from the 28 nm technology node to the 5 nm technology node, the SRAM cell area according to the present invention is 84λ2 ~102λ 2 Taking a technology node or minimum feature size = 5 nm as an example, the area of ​​the newly proposed SRAM cell is 100λ, which is almost one eighth (1 / 8) of the area of ​​the conventional 5 nm SRAM cell shown in Figure 32. 2 It may be to some extent.

[0089] Of course, it is not necessary to utilize all of the proposed improvements in the novel SRAM cell structure of the present invention, and only one of the proposed techniques can reduce the area of ​​the SRAM cell structure compared to a transitional SRAM cell. For example, by reducing the active area (or simply connecting the gate / source / drain contacts ("CT") with a second metal layer) in accordance with the present invention, the area of ​​the SRAM can be reduced by 84λ at the 5 nm technology node. 2 ~700λ 2 In the range of 84λ for the 7nm technology node 2 ~450λ 2 Within the range of 10nm to 7nm technology nodes, 84λ 2 ~280λ 2 Within the range of technology nodes from 20nm to over 10nm, 84λ 2 ~200λ 2 Within the range of technology nodes from 28nm to over 20nm, 84λ 2 ~150λ 2 For example, by reducing the area of ​​the active region, the area of ​​the SRAM can be reduced to 160λ at ​​the 5 nm technology node. 2 ~240λ 2 (or more if further tolerances are required) and the area of ​​the SRAM is 107λ 2 ~161λ 2 (or more if further tolerance is required).

[0090] The area (λ 2), the linear dimensions of the present invention may be 0.9 (or less, e.g., 0.85, 0.8, or 0.7) times the linear dimensions of the conventional SRAM of FIG. 3, and the area of ​​the present invention may be at least 0.81 (or less, e.g., 0.72, 0.64, or 0.5) times the area of ​​the conventional SRAM of FIG. 2B. Thus, in another embodiment, when the minimum feature size is 5 nm, the area of ​​the SRAM cell is 84λ 2 ~672λ 2 When the minimum feature size is 7 nm, the area of ​​the SRAM cell is 84λ 2 ~440λ 2 When the minimum feature size is between 10 nm and over 7 nm, the area of ​​the SRAM cell is 84λ 2 ~300λ 2 When the minimum feature size is between 16 nm and over 10 nm, the area of ​​the SRAM cell is 84λ 2 ~204λ 2 When the minimum feature size is between 22 nm and over 16 nm, the area of ​​the SRAM cell is 84λ 2 ~152λ 2 When the minimum feature size is between 28 nm and over 22 nm, the area of ​​the SRAM cell is 84λ 2 ~139λ 2 is within the range.

[0091] Similarly, logic circuits based on standard cells can be applied with the above transistor, CMOS, latch-up designs, and / or wiring structures. In Figures 33A and 33B, a new standard cell (cell area: 2Cpp x Cell_Height) is proposed, where Cpp can be as small as 4λ and Cell_Height can be as small as 24λ. Note that in Figure 33A, two active fins are used for PMOS and NMOS, respectively. However, the fin pitch can be as small as 3λ. The width of the active area or fin is as small as λ, as is the width of the gate line (or polyline). Its dimensions can be easily formed regardless of the size of the currently available technology node (or minimum feature size). Thus, the cell area of ​​the proposed standard cell is (2Cpp×Cell_Height) is 192λ 2 It is.

[0092] As shown in FIG. 33B, several source / drain contacts (for connecting to the metal 1 layer) can be formed in the active area. The size of the source / drain contacts can be as small as λ×λ, regardless of the technology node or (or minimum feature) size. Similarly, gate contacts (for connecting directly to the metal 2 layer without connecting the metal 1 layer as explained above) can be formed on the gate or poly line, and the size of the gate contacts can be as small as λ×λ. That is, the source, drain, gate and their contacts of the transistors in the standard cell can have precisely controlled linear dimensions and can be as small as lambda (λ), which is the minimum feature. In this embodiment, the spacing between two gate or poly lines is as small as 3λ.

[0093] Furthermore, since the bottom surface of the source / drain structure can be isolated from the substrate as described above, the n+ to n+ or p+ to p+ insulation can be kept within a moderate range. Therefore, the distance between two adjacent active regions can be reduced to 2λ (shown by the dotted circle on the left of FIG. 33B). Furthermore, in the present invention, since the n+ and p+ regions of the source and drain regions in the NMOS and PMOS transistors are completely insulated by an insulator, respectively, it is possible to reduce the latch-up distance between PMOS and NMOS to 8λ (shown by the dotted line on the right of FIG. 33B) regardless of the size or (minimum feature) of the technology node.

[0094] As a result, according to the present invention, the area size of a standard cell (2Cpp×Cell_Height) that can accommodate an inverter is 192λ 2 And such a λ 2The area size in units of Cpp is almost the same at least from technology node 22 nm to 5 nm, as shown in Figure 34A. Compared with the conventional performance in other semiconductor manufacturing plants, the proposed standard cell (2 Cpp × cell height) has an area of ​​about 1 / 3.5 of the conventional 5 nm standard cell.

[0095] Of course, it is not necessary to utilize all of the proposed improvements in the novel standard cell of the present invention, and only one of the proposed techniques can reduce the area of ​​the standard cell structure compared to the transitional standard cell. For example, the area (2Cpp x cell height) of the standard cell of the present invention is 190λ at ​​the 5 nm technology node. 2 ~600λ 2 in the range of 190λ at ​​the 7nm technology node 2 ~450λ 2 or 190λ 2 ~250λ 2 190λ in technology nodes between 10nm and 14nm 2 ~250λ 2 It can be within the range.

[0096] Furthermore, in another embodiment, the present invention can be used with different cell sizes such as 3Cpp×Cell_Height or 5Cpp×Cell_Height. Two NOR cells, two NAND cells, or two inverter cells may be embedded in a cell size of 3Cpp×Cell_Height, and two NOR cells or two NAND cells may be embedded in a cell size of 5Cpp×Cell_Height. Also, the area size of the proposed standard cell (regardless of whether the cell size is 3Cpp×Cell_Height or 5Cpp×Cell_Height) is λ 2 It can be concluded that the MOSFETs are almost the same at least from technology node 22nm to 5nm.

[0097] Figure 34B shows the Cpp, fin pitch, and Cell_Height values ​​for different technology nodes from three different semiconductor foundries and the present invention implemented with some of the proposed transistor structures and wiring with extra tolerances. The Cpp and fin pitch values ​​of the present invention are applicable not only to SRAM cells but also to standard cells (as shown in Figures 31B and 33A). Of course, it is not necessary to utilize all the proposed improvement techniques in the new die, and only one of the proposed techniques can reduce the area of ​​the SRAM cell or standard cell structure compared to the transitional SRAM cell. Thus, compared to other available semiconductor foundries, the value of Cpp according to the present invention may be 45 nm or less (e.g., in the range of 45-20 nm or 40-20 nm) at the 5 nm technology node, 50 nm or less (e.g., in the range of 50-28 nm or 45-28 nm) at the 7 nm technology node, 50 nm or less (e.g., in the range of 50-40 nm or 45-40 nm) at the 10 nm technology node, and 67 nm or less (e.g., in the range of 67-64 nm) at the technology node between 14 nm and 16 nm. Furthermore, the value of fin pitch according to the present invention may be 20 nm or less (e.g., in the range of 20 nm-15 nm) at the 5 nm technology node, 24 nm or less (e.g., in the range of 24 nm-21 nm) at the 7 nm technology node, and 32 nm or less (e.g., in the range of 32 nm-30 nm) at the 10 nm technology node.

[0098] Furthermore, the value of Cpp can be 45 nm or less (e.g., in a range of 45 nm to 20 nm) when the second fin width is 5 nm or less, or the value of Cpp can be 50 nm or less (e.g., in a range of 50 nm to 28 nm) when the second fin width is 7 nm or less and 5 nm or more, or the value of Cpp can be 50 nm or less when the second fin width is 10 nm or less and 7 nm or more. (eg, in the range of 50 nm to 40 nm), or the value of Cpp may be 67 nm or less (eg, in the range of 67 nm to 64 nm) when the second fin width is 14 nm to 16 nm.

[0099] According to the above, FIG. 35 discloses the present innovation of Integrated Scaling and / or Stretching Platform (ISSP) in its monolithic die design. First, the original circuit of die A is reduced in area by 2-3 times due to the proposed novel transistor, CMOS, wiring structure, etc., so that the main functional blocks such as CPU or GPU can be shrunk to a significantly smaller size. Then, more SRAM or main functional blocks can be formed on a single monolithic die. Taking the 5nm technology node as an example, the SRAM cell size of 6T can be reduced to about 100F as shown in FIG. 32. 2 (F is the smallest feature that can be made on a silicon wafer). In other words, if F=5nm, the SRAM cell will be about 2500nm 2 The state-of-the-art cell area is about 800F based on the paper. 2 (~8x reduction). In addition, an 8-finger CMOS inverter (shown in Figures 33A and 33B with dimensions 2Cpp x Cell_Height) can be fabricated at 700F in the 5nm process node of Figure 34A. 2 Over or 900F 2 In contrast to published CMOS inverters up to 200F 2 of die area must be consumed.

[0100] That is, die A has a first die area (Ynm 2 In the case where a die A has a circuit (SRAM circuit, logic circuit, combination of SRAM+logic circuit, or major functional block circuit CPU, GPU, FPGA, etc.) that occupies a large amount of space (e.g., Ynm), with the help of the present invention, the total area of ​​die A with the same circuit can be reduced even if die A is still manufactured by the same technology node. Moreover, the new die area occupied by the same circuit on die A will be smaller than the original die area, e.g., Ynm 2 The ratio is 20% to 90% (or 30% to 70%) of the total.

[0101] For example, as shown in Figure 35, the original SOC die 3510 is 26x33mm 2 The scanner maximum exposure area (SMFA) of the die is 26x33mm, of which the original SRAM, original logic, and I / O pads occupy 65%, 25%, and 10% of the die area, respectively. If the SRAM is shrunk by 5.3 times and the logic is shrunk by 3.5 times, the new shrunk die 3520 is 26x33mm 2 This is 1 / 3.4 the die area of ​​the SMFA. Therefore, the die area is 26×33mm 2 In another aspect, based on the proposed Integrated Scaling and / or Stretching Platform (ISSP), it is easily possible to combine more SRAM (e.g., 5.7 times the amount of original SRAM) with the shrunk die 3520 in the same SMFA into a new monolithic die 3530, or combine more major functional blocks (e.g., a new CPU, a new GPU, a new FPGA, etc.) with the shrunk die 3520 in the same SMFA into another new monolithic die 3530.

[0102] Thus, more SRAMs are formed on the monolithic die. Currently, major processing units (such as CPUs or GPUs) have several levels of caches. L1 and L2 caches (collectively called "low level caches") are usually one per core unit of a CPU or GPU, with the L1 cache divided into L1i and L1d used to store instructions and data, respectively, the L2 cache that does not distinguish between instructions and data, and the L3 cache (which may be one of the "high level caches") that is shared by multiple cores and usually does not distinguish between instructions and data either. There is usually one L1 / L2 cache per core of a CPU or GPU, which means that an area of ​​the same size must be increased for each additional core of the CPU or GPU. Usually, the more the capacity of the cache, the higher the hit rate. For high speed operation, these low level or high level caches are generally composed of SRAM. Therefore, based on our integrated scaling and / or stretching platform, the L1 / L2 caches ("low level caches") and L3 caches ("high level caches") can be increased within a monolithic single die within the Scanner Maximum Field Area (SMFA) limited by the photolithography exposure tool.

[0103] In one example, as shown in FIG. 36A for a single monolithic die 3610, an XPU 3610 (e.g., a GPU) with multiple cores may have a SMFA (e.g., 26 mm×33 mm, or 858 mm×128 mm) with high level caches that may have 64 MB of SRAM or more (e.g., 128, 256, 512 MB or more). 2In addition, additional logical GU cores (GU core 1 to GU core 2N, e.g., 64, 128, 256 or more cores) of the GPU can be inserted in the same SMFA to improve performance. In another embodiment, the same is true for a memory controller with high bandwidth input / output. Each monolithic die includes an input / output bus (e.g., high bandwidth input / output), and each CPU or GPU core is electrically coupled to the input / output bus, and their caches and SRAMs are also electrically coupled to the input / output bus.

[0104] Alternatively, other major functional blocks than the completed major functional blocks, such as a network processing unit (NPU), a tensor processing unit (TPU) or an FPGA, which are also reduced according to the present invention, can be integrated together in another monolithic die 3620 as shown in FIG. 36B. The XPU3621 and YPU3622 in FIG. 36B represent processing units having major functional blocks, which can be NPU, GPU, CPU, FPGA, or TPU. For example, the XPU3621 can be a CPU and the YPU3622 can be a GPU. The major functional blocks of the XPU3621 can be the same as or different from those of the YPU3622. The XPU3621 and YPU3622 have multiple logical cores, each with its own low level cache (e.g. L1 / L2 cache - 128K for L1, 512K or 1M for L2), and a larger higher level cache (e.g. 32MB, 64MB or more for L3 cache) shared between the XPU3621 and YPU3622. Each monolithic die includes an I / O bus (e.g. high bandwidth I / O), and each logical core is electrically coupled to the I / O bus, and their cache or SRAM is also electrically coupled to the I / O bus.

[0105] Thus, a single monolithic die of the present invention (which may have a scanner maximum exposure area) may have two (or more) major functional blocks or different circuits. A conventional monolithic die may have a first circuit or a first major functional block occupying 20%-90%, 30%-80%, 50%-90% or 60%-90% of the scanner maximum exposure area of ​​the conventional monolithic die (e.g., as shown on the left side of FIG. 35, a logic circuit corresponding to a circuit may occupy about 25%-30%, an SRAM circuit corresponding to a circuit may occupy about 50%-65%, and a combination of SRAM and logic circuit corresponding to another circuit may occupy about 80%-90%). However, a single monolithic die of the present invention (i.e., made based on the same technology node as the conventional monolithic die, such as 5 nm or 7 nm) having the same scanner maximum exposure area may not only include the same first circuit or first major functional block, but may also include another second circuit or second major functional block (as shown on the right side of FIG. 35). In another embodiment, the area of ​​the second circuitry in the monolithic die of the present invention is similar to the area of ​​the first circuitry in the monolithic die of the present invention.

[0106] According to the present invention, the first circuit or the first main functional block in a conventional monolithic die can be reduced by 20% to 90% (for example, 30% to 80%, in FIG. 32 and FIG. 34A, the SRAM circuit can be reduced to 1 / 8, and the logic circuit can be reduced to 1 / 3.5). In particular, GPUs are more suitable for AI learning, but are not suitable for AI inference. On the other hand, FPGAs have blocks of interacting logic that engineers can design to assist specific algorithms, and are suitable for AI inference. Based on an integrated scaling and / or stretching platform (ISSP), both GPUs and FPGAs may be formed in a monolithic die. Such a monolithic die is excellent in parallel processing, learning speed, and efficiency on the one hand. On the other hand, it also possesses excellent AI inference capabilities with short time to market, low cost, and high flexibility.

[0107] In another embodiment, as shown in FIG. 36C for a single monolithic die 3630, a shared high level cache 3633 (e.g., L3 cache) between the XPU 3631 and the YPU 3632 is configurable by settings in a separate mode register (not shown) or by being adaptively configurable during operation of the monolithic die. For example, in one embodiment, by setting a mode register, 1 / 3 of the high level cache 3633 can be used by the XPU 3631 and 2 / 3 of the high level cache 3633 can be used by the YPU 3632. The amount of such sharing of high level cache 3633 (e.g., L3 cache) for the XPU 3631 or YPU 3632 can also be changed dramatically based on the operation of the Unified Scaling and / or Stretching Platform (ISSP). In yet another embodiment, as shown in Figure 36D for a single monolithic die 3640, the higher level caches include an L3 cache 3643 and an L4 cache 3644, with each of the XPU 3641 and YPU 3642 having a corresponding L3 cache (e.g., 8M or more) 3643 shared by its own cores, and an L4 cache 3644 (e.g., 32MB or more) shared by the XPU 3641 and YPU 3642. Again, in this example, each monolithic die includes an I / O bus (such as a high bandwidth I / O), with each logic core electrically coupled to the I / O bus, as well as their caches or SRAMs electrically coupled to the I / O bus.

[0108] Of particular importance is that due to the much smaller area of ​​the eSRAM design according to the present invention, a somewhat larger capacity shared SRAM (or embedded SRAM, "eSRAM") can be designed into the die. Since more and smarter shared eSRAMs can be used, it may be more effective to connect an external DRAM to this eSRAM in a monolithic die with a limited SMFA corresponding to a particular technology node, and the capacity of the external DRAM required may be reduced. Thus, the present invention discloses a platform for reconfiguring the memory architecture of a conventional chip system. In a conventional chip system, the first monolithic die (e.g., GPU) is connected to a first DRAM memory having a first predetermined capacity (e.g., 1 GB), the first monolithic die has a scanner maximum exposure area (SMFA) based on a targeted technology node (e.g., 5 nm), and includes a first logic circuit and a first SRAM memory, and the sum of the area of ​​the first logic circuit and the area of ​​the first SRAM memory occupies at least 80-90% of the scanner maximum exposure area of ​​the first monolithic die.

[0109] Based on the present invention, a new platform with a completely new memory architecture includes a second monolithic die connected to a second DRAM memory, the second monolithic die includes a second logic circuit and a second SRAM memory, and the second monolithic die has the same SMFA based on the same targeted technology node. Here, the second logic circuit is identical or substantially identical to the first logic circuit (e.g., both are all the same GPU or NPU), but the area of ​​the first logic circuit in the first monolithic die is larger than the area of ​​the second logic circuit in the second monolithic die because the area of ​​the second logic circuit can be reduced according to the present invention. Furthermore, because the area of ​​the SRAM structure according to the present invention is very small and there is a lot of spare chip area left on the second monolithic die, a larger capacity SRAM can be designed into the die, and thus the capacity of the second SRAM memory in the second monolithic die is much higher than the capacity of the first SRAM memory in the first monolithic die. Because the capacity of the SRAM of the second monolithic die is large, the capacity of the DRAM connected to the second monolithic die can be reduced so that the capacity of the second DRAM memory is smaller than the capacity of the first DRAM memory.

[0110] In another embodiment shown in FIG. 36E for a single monolithic die 3650, one good approach is to have a single large Direct Wide Bus (DWB) on the monolithic die (scalable up to the maximum reticle size allowed) connected to another monolithic die of external DRAM or other embedded DRAM ("eDRAM"). The DWB was filed on June 18, 2020, and is incorporated herein by reference in its entirety for the purposes of the present application. No. 16 / 904,597, entitled "DWB for Data Linking Devices and Devices with a Chip," the entire contents of which are incorporated herein by reference. The DWB can have 128 bits, 256 bits, 512 bits, 1024 bits, or more bits and transmit the date in parallel. In FIG. 36E, embedded DRAM ("eDRAM") 3656 may be located in a separate die packaged with monolithic die 3650 having at least two major functional blocks (XPU 3651 and YPU 3652) and large capacity SRAM (e.g., L3 cache 3653 and L4 cache 3654, etc.). External DRAM 3657 is separate from package 3655 but communicates with single monolithic die 3650 through the DWB. Additionally, single monolithic die 3650 with limited SMFA corresponding to a particular technology node also includes a memory controller and physical layer compatible with the DWB.

[0111] In summary, the monolithic / heterogeneous integration on a single die that enabled the success of Moore's Law is currently facing limitations, especially due to the limitations of photolithography technology. On the one hand, the smallest feature printed on the die is very costly to scale to its dimensions, and on the other hand, the die size is limited by the scanner's maximum exposure area. However, the increasing and diversifying functions of processing equipment make it difficult to integrate them on a monolithic die. Moreover, the somewhat overlapping presence of eSRAM on the die for each major function and an external DRAM that serves only each individual die function is not an optimized and desirable solution. Based on the proposed Integrated Scaling and / or Stretching Platform (ISSP) on a monolithic or SOC die, (a) a single major functional block such as FPGA, TPU, NPU, CPU, or GPU can be shrunk to a significantly smaller size, (b) more SRAM or more functional blocks can be formed on a monolithic die, (c) the GPU and FPGA can be integrated into a single die. Two or more major functional blocks that are also scaled through this ISSP, such as (or other combinations), can be integrated together on the same monolithic die. Additionally, there can be more levels of cache within the monolithic die. Such an integrated monolithic die can be combined with another die based on heterogeneous integration (e.g. eDRAM).

[0112] Although the invention has been described with reference to exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims. [Explanation of symbols]

[0113] 100 mm MOSFET 101 Gate Structure 102 p type substrate 103 Source 104 Source / Drain Region 105 Quarantine area 107 Drain 109, 111 Contact Hole 120 First dielectric layer 150 First conductive layer 160 Second dielectric underlayer 170 Upper dielectric layer 180 Upper conductive layer 302 Pad Oxide Layer 304 Pad nitride layer 402 Dielectric Insulators 602 Gate Layer 604 Nitride layer 702 Spin-on Dielectric 802 Gate Mask Layer 1502 Oxide 3 Spacer 1504 Lightly doped drain 1602 Intrinsic silicon electrode 1702 CVD-STI oxide 3 layer 1704 Source Region 1706 Drain Region 1802 Oxide Spacer 1901 SOD 1902 Photoresistance layer 1901 SOD layer 1903 Gate Extension Area

Claims

1. A monolithic die, a first processor circuit formed on the monolithic die having a scanner maximum exposure area, the first processor circuit comprising a plurality of first logic cores, each first logic core corresponding to a first cache memory; a second processor circuit formed on the monolithic die, the second processor circuit comprising a plurality of second logic cores, each second logic core corresponding to a second cache memory; A monolithic die, wherein the scanner maximum exposure area of ​​the monolithic die is defined by a particular technology node.

2. The scanner maximum exposure area of ​​the monolithic die is 26 mm x 33 mm or 858 mm 2 2. The monolithic die of claim 1 , wherein:

3. The monolithic die of claim 1 , wherein a primary function performed by said first processor circuitry is different from a primary function performed by said second processor circuitry.

4. The monolithic die of claim 3 , wherein the first processor circuit or the second processor circuit is selected from the group consisting of a GPU, a CPU, a TPU, an NPU, and an FPGA.

5. 2. The monolithic die of claim 1, further comprising a third cache memory, said third cache memory being configurable and utilized by said first processing unit circuitry and said second processing unit circuitry during operation of said monolithic die, said first cache memory, said second cache memory, and said third cache memory being comprised of SRAM.

6. 2. The monolithic die of claim 1, further comprising a third processing unit circuit formed on said monolithic die, said third processing unit circuit comprising a plurality of third logic cores, each third logic core corresponding to a third cache memory.

7. A monolithic die, a first processor circuit formed on the monolithic die having a scanner maximum exposure area, the first processor circuit comprising a plurality of first logic cores, each first logic core corresponding to a first cache memory; a second processor circuit formed on the monolithic die, the second processor circuit comprising a plurality of second logic cores, each second logic core corresponding to a second cache memory; a third cache memory shared and utilized by the first processor circuitry and the second processor circuitry during operation of the monolithic die; A monolithic die, wherein the primary function performed by the first processor circuitry is different from the primary function performed by the second processor circuitry.

8. 8. The monolithic die of claim 7, wherein the first processor circuit or the second processor circuit is selected from the group consisting of a GPU, a CPU, a TPU, an NPU, and an FPGA.

9. 8. The monolithic die of claim 7, wherein the first cache memory, the second cache memory, and the third cache memory are comprised of SRAM.

10. A monolithic die, a first processor circuit formed on the monolithic die, the first processor circuit comprising a plurality of first logic cores, each first logic core circuit corresponding to a lower level cache memory; a high level cache memory utilized by the first processing unit circuit; a monolithic die, the sum of the size of all of said lower level cache memories and the size of said higher level cache memories being at least 128 MB;

11. 11. The monolithic die of claim 10, wherein a scanner maximum exposure field area of ​​the monolithic die is defined by a photolithography exposure tool utilized for a first technology node.

12. The scanner maximum exposure area of ​​the monolithic die is 26 mm x 33 mm or 858 mm 2 11. The monolithic die of claim 10, wherein:

13. The monolithic die of claim 10 , wherein the plurality of first logic cores comprises at least 128 first logic cores.

14. 11. The monolithic die of claim 10, wherein the lower level cache memory includes an L1 cache and an L2 cache, the size of the L2 cache being greater than the size of the L1 cache, and the higher level cache memory is shared and utilized by the plurality of first logic core circuits.

15. 15. The monolithic die of claim 14, further comprising a second processor circuit formed on the monolithic die, the second processor circuit comprising a plurality of second logic core circuits, each second logic core circuit corresponding to a second cache memory, the higher level cache memory including an L4 cache shared and utilized by the first processor circuit and the second processor circuit.

16. 16. The monolithic die of claim 15, wherein the L4 cache is shared and utilized by the first processing unit circuitry and the second processing unit circuitry depending on a mode register setting.

17. 1. A method for manufacturing a monolithic die having a set of CMOS circuitry, comprising the steps of: performing a first predetermined process based on a first technology node to control dimensions of a plurality of transistors in the set of CMOS circuits of the monolithic die; performing a second predetermined process based on the first technology node to interconnect the transistors of the set of CMOS circuits to other metal layers in the monolithic die; a new size of the monolithic die manufactured by the first predetermined process and the second predetermined process is smaller than an original size of another monolithic die having the same set of CMOS circuits manufactured based on the first technology node without performing the first predetermined process and the second predetermined process.

18. 20. The method of claim 17, wherein the new size of the monolithic die produced by the first predetermined process and the second predetermined process is less than 50% of the original size of the another monolithic die.

19. 20. The method of claim 17, wherein the new size of the monolithic die manufactured by the first predetermined process and the second predetermined process is less than 35% of the original size of the another monolithic die when the first technology node is 5 nm.

20. 1. A method for manufacturing a monolithic die having a plurality of SRAM cells, comprising: performing a first predetermined process based on a first technology node to control dimensions of a plurality of transistors of an SRAM cell in the monolithic die; performing a second predetermined process based on the first technology node to interconnect the transistors of an SRAM cell to other metal layers in the monolithic die; A manufacturing method, wherein a total capacity of the plurality of SRAM cells of the monolithic die manufactured by the first predetermined process and / or the second predetermined process is greater than a total capacity of another monolithic die manufactured based on the first technology node without performing the first predetermined process and the second predetermined process.

21. 21. The manufacturing method of claim 20, wherein the monolithic die having a scanner maximum exposure area area defined by the first technology node has a first processing device circuit and a second processing device circuit, a primary function performed by the first processing device circuit is different from a primary function performed by the second processing device circuit, and a total capacity of the plurality of SRAM cells in the monolithic die manufactured by the first predetermined process and / or the second predetermined process is at least 128 MB.

22. A first monolithic die, a first circuit fabricated by performing a predetermined manufacturing process based on a first technology node; A first monolithic die having a die area smaller than a die area of ​​a second monolithic die comprising a second circuit fabricated based on the first technology node without performing the predetermined manufacturing process.

23. 23. The first monolithic die of claim 22, wherein the first circuit is the same as the second circuit.

24. 24. The first monolithic die of claim 23, wherein the second circuitry occupies between 20% and 90% of the die area of ​​the second monolithic die.

25. The first circuit is a Ynm of the first monolithic die. 2 and the second circuit occupies Xnm of the second monolithic die. 2 25. The first monolithic die of claim 24, wherein X>Y.

26. The first monolithic die of claim 25, wherein Y is between 20% and 90% of X.

27. 24. The first monolithic die of claim 23, wherein the first circuit is an SRAM circuit, a logic circuit, a combination of SRAM and logic circuit, or a major functional block circuit.

28. A first monolithic die, a first circuit fabricated according to a first technology node; a die area of ​​the first monolithic die is smaller than a die area of ​​a second monolithic die having a second circuit fabricated based on the first technology node, the first circuit being the same as the second circuit; A first monolithic die, wherein the first circuit is an SRAM circuit, a logic circuit, a combination of an SRAM and a logic circuit, or a major functional block circuit.

29. 30. The first monolithic die of claim 28, wherein the second circuitry occupies between 20% and 90% of the die area of ​​the second monolithic die.

30. The first circuit is Ynm of the first monolithic die. 2 and the second circuit occupies Xnm of the second monolithic die. 2 30. The first monolithic die of claim 28, wherein X>Y.

31. 30. The first monolithic die of claim 29, wherein Y is between 20% and 90% of X.

32. A first monolithic die, a first circuit formed on the first monolithic die; a second circuit formed on the first monolithic die; the first monolithic die having a first scanner maximum exposure area based on a first technology node, the first circuitry occupying a first portion of the first scanner maximum exposure area, and the second circuitry occupying a second portion of the first scanner maximum exposure area; A first monolithic die, wherein the scanner maximum exposure area of ​​the first monolithic die is the same as a second scanner maximum exposure area of ​​a second monolithic die, the second monolithic die having the first circuit, and an area of ​​the first circuit of the second monolithic die is between 50% and 90% of the second scanner maximum exposure area of ​​the second monolithic die.

33. 33. The first monolithic die of claim 32, wherein the first circuit is an SRAM circuit, a combination of an SRAM circuit and a logic circuit, or a major functional block circuit.

34. 33. The first monolithic die of claim 32, wherein the first circuit is selected from the group consisting of a GPU, a CPU, a TPU, an NPU, and an FPGA.

35. The first scanner maximum exposure area or the second scanner maximum exposure area is 26 mm x 33 mm or 858 mm 2 33. The first monolithic die of claim 32, wherein:

36. A first monolithic die, an SRAM circuit including an SRAM cell, the SRAM cell having a first fin pitch and a first contact poly pitch; a logic circuit including a standard cell electrically coupled to the SRAM circuit, the standard cell including a second fin pitch and a second contact poly pitch. a value of the first contact poly pitch or the second contact poly pitch is equal to or less than 45 nm at a 5 nm technology node; or a value of the first contact poly pitch or the second contact poly pitch is less than or equal to 50 nm at a 7 nm technology node; or a value of the first contact poly pitch or the second contact poly pitch is 50 nm or less at a 10 nm technology node; or A first monolithic die, wherein the first contact poly pitch or the second contact poly pitch has a value of 67 nm or less in a 14-16 nm technology node.

37. A first monolithic die, an SRAM circuit including an SRAM cell, the SRAM cell having a first fin pitch and a first contact poly pitch; a logic circuit including a standard cell electrically coupled to the SRAM circuit, the standard cell including a second fin pitch and a second contact poly pitch. the value of the first fin pitch or the second fin pitch is less than or equal to 20 nm at a 5 nm technology node; or the value of the first fin pitch or the second fin pitch is less than or equal to 24 nm at a 7 nm technology node; or A first monolithic die, wherein a value of the first fin pitch or the second fin pitch is less than or equal to 32 nm at a 10 nm technology node.

38. A first monolithic die, an SRAM circuit including an SRAM cell, the SRAM cell having a first fin pitch, a first fin width, and a first contact poly pitch; a logic circuit including a standard cell electrically coupled to the SRAM circuit, the standard cell including a second fin pitch, a second fin width, and a second contact poly pitch; When the second fin width is 5 nm or less, the value of the second contact poly pitch is in the range of 45 nm to 20 nm; or When the second fin width is 7 nm or less and 5 nm or more, the value of the second contact poly pitch is in the range of 50 nm to 28 nm; or When the second fin width is 10 nm or less and 7 nm or more, the value of the second contact poly pitch is in the range of 50 nm to 40 nm; or The first monolithic die, wherein the second contact poly pitch has a value in the range of 67 nm to 64 nm when the second fin width is 14 nm to 16 nm.

39. 1. A platform for reconfiguring a memory architecture of a chip system, the chip system including a first monolithic die coupled to a first DRAM memory having a first predetermined capacity, the first monolithic die including a first logic circuit and a first SRAM memory, the platform comprising: a second monolithic die coupled to a second DRAM memory having a second predetermined capacity, the second monolithic die including a second logic circuit and a second SRAM memory; the first monolithic die having a scanner maximum exposure area based on a target technology node, and the second monolithic die having the scanner maximum exposure area based on the target technology node; the first logic circuit is the same as the second logic circuit, and an area of ​​the first logic circuit on the first monolithic die is greater than an area of ​​the second logic circuit on the second monolithic die; A platform, wherein a capacity of the first SRAM memory is less than a capacity of the second SRAM memory, and a second predetermined capacity of the second DRAM memory is less than a first predetermined capacity of the first DRAM memory.

40. 40. The platform of claim 39, wherein a sum of an area of ​​the first logic circuitry and an area of ​​the first SRAM memory occupies at least 80% of a scanner maximum exposure area area of ​​the first monolithic die, and a sum of an area of ​​the second logic circuitry and an area of ​​the second SRAM memory occupies at least 80% of a scanner maximum exposure area area of ​​the second monolithic die.

41. 40. The platform of claim 39, wherein the second DRAM memory is an embedded DRAM chip external to the second monolithic die, and wherein the second DRAM memory and the second monolithic die are enclosed in a single package.

Citation Information

Patent Citations

  • buffer memory management system for automation device

    DE102017203378A1

  • Semiconductor integrated circuit and designing method for the same

    JP2003337760A

  • Semiconductor device

    JP2006165480A

  • Monolithic integrated circuit die having modular die regions stitched together

    JP2016530704A

  • Cache memory device and semiconductor device

    JP2018106227A