This slide shows the semiconductor fabrication. The purpose of this class is learning about MOSFET, and if you don't know how to fabricate MOSFET, you probably don't understand clearly about the theory of a semiconductor. But this semiconductor fabrication is separated class you can learn in other lectures. First, semiconductor starting with a n-type silicon wafer. This can be started with a p-type semiconductor, but this is the n-type semiconductor. We glow silicon oxide in high temperature furnace with oxygen gas. Then using photolithography with a photoresist spin-coating of the polymer and exposing the mask of the UV light, then you can pattern the photoresists on top of the silicon oxide. Then you put this material in etching solution or plasma etching, then you can transfer pattern of a PR to silicon oxide. Then using solid-state doping on ion implantation, you dope source drain region with a highly dopant. Then you glow silicon gate directory oxide again and part on the contact area of source drain region and depositing the metal. Using PI photolithography, you're patterning the metal source drain gate. For more detailed information, I recommend you to take other semiconductor fabrication classes. Many of you are interested in going on industry of a semiconductor. If you go to the semiconductor industry, this is the four different areas that I can define. First, device engineer and integration engineer, process engineer, and TCAD engineer. Device engineer is designing the semiconductor architects. Let's say that I want to design the smallest semiconductor device as shown in this image. Then we have to make decision related with the target performance, device structure, notch or polysilicon or what kind of junction we're going to use, and material for the gate material and oxide material, and what kind of doping, how much doping we are going to do and we making up process flow chart, each defining each process. Integration engineer is handling the entire semiconductor process, moving the silicon lab that has a lot of silicon wafer. One process to the process following the process flow chart, and making a mask layout, and optimization of the device condition using spirit of the channel doping, strain engineering, junction technology, etc. Process engineer take charge of the individual process, mostly done by the material scientists such as the plasma etching, lithography, electronic material of thin-film, TEM material analysis, ion implantation, metallization, and packaging. Semiconductor industry is divided by the two areas of the logic and memory. Logic is also called non-memory system semiconductor. Non-memory device or logic computer is the largest semiconductor, is 70 percent of the total semiconductor market, very huge. Including this spray driver IC, DDI, Graphic Processing Unit, GPU, navigation processor, CMOS image sensor, which is using your digital camera, and CPU, also MPU, and Application Specific Integrated Circuit and Digital Signal Processing. This non-memory area has various products and high profit. For the company such as the Qualcomm and ARM doesn't have the Fab, and they're designing the logic semiconductor device called the Fabless company. If the Qualcomm and ARM or Fabless company design the semiconductor, those example of the TSMC, which is the Taiwan Semiconductor Manufacturing Company, which is the largest founder in the world making those non-memory device. This non-memory device has a complicated circuit such as the ALU, Arithmetic Logic Unit, and CU, Control Unit, and SRAM of the memory and input and output. If you look at the live image of the Von Neumann Architecture, this is how the CPU works. By the Von Neumann's architecture, consists with those CU, ALU, memory, input, and output. If you look at the computer, there is a core of a CU and ALU and there is the cache memory, which is the SRAM, and the main memory of the RAM, which may be DRAM and flash memory. Cache memory is included in CPU and DRAM and flash memory is separated devices. Memory industry has also big size. Memory can be set DRAM of the main memory of a huge size of the memory. SRAM, as I said, is inside the CPU, which is very high speed and high cost. What's the difference between the SRAM and DRAM? Let's look at the DRAM structure. The circuit is the below image that is consisted with one MOSFET device, the word line connected to the gate of the MOSFET and bit line is connected to the one end of the source string region. The select transistor is the MOSFET and the other end of the MOSFET connected to the capacitor where charge is stored. When charge is stored, data 1, charge is not stored, data 0. That's the DRAM operation mechanism. This is to separate the silicon chip. But SRAM cases is located inside the CPU and circuit structure is the light upper images showing that consisted with a six MOSFET transistor. This six transistor can store the data of 1 or 0. So if you look at the occupying area of SRAM and DRAM, maybe six times more occupy the area. If you have a high occupying area, then cost is high. However, they are located inside the CPU, then the speed is very fast. Another main memory is the non-volatile flash memory. Volatile and non-volatile means that for DRAM, if you turn up the switch, your memory data disappear. Non-volatile, like a flash memory, store the data even if you turn off the switch. Your computer probably have a hard disk that's a magnetic-based memory device. However, very vastly changing to the solid state drive of the flash memory, especially in the laptop, to reduce the size and reduce the light weight and fast and quiet operations. Currently, Samsung Electronics, SK Hynix, Toshiba, and Micron Technology make main memory devices. Other technology may be called the SOC, system on chip is having a whole CPU and main memory, input and output in one single chip, which is the very fastest speed. If you think about the SOC is the most ideal case, however, it's not, because to making a CPU and memory in one chip is very difficult to fabricate, because they have a different structures. Let's look at the semiconductor market. When semiconductor market is open, US companies dominated. However, late 1980, Japanese company is dominating the semiconductor market. If you look at the 20 company; NEC, Toshiba, Hitachi, Fujitsu, Matsushita, Mitshubishi, Sanyo, Sharp, Sony, etc, is all Japanese-based company. However, the right images shows the semiconductor market in 2018, it's a big change. First company is the Samsung, and second, Intel, third, SK Hynix, TSMC, Micorn, and so on. So only Toshiba of Japanese company is survived this [inaudible]. What about the future? It depends on the semiconductor engineer, it's you, taking these classes. Finally, where this Silicon technology heading for? I taught this class for last 11 years in KAIST. Sometimes students said, "Silicon technology is saturating, it will be end soon." So they should study other research field. But I don't think so. I think Silicon technology is changing from the PC era, and mobile era, and they shifting the paradigm of fusion era, such as integrating lab on a chip and bio application, medical device, everything in one single fusion technology. Also Silicon technology is the driving force of the artificial intelligent and big data, Future car industry, automobile industry will depend on the semiconductor, because electric car and inside the electric car, everything is digital devices. Our future society will be based on the IoT, Internet of Thing. Every object becomes a computer, then these Silicon technology will be exploding again. The left below image is the yearly data production. In 2025, the data we will use is the 165 zettabyte. Zetta is the 10 to the 21 so 100 zettabyte is 10 to the 23. If you learn the chemistry, you are familiar with this 10 to the 23, which is almost equal to the Avogadro numbers. To prepare this huge data society, semiconductor industry has to support this future technology. Another important field is the TCAD engineering. TCAD is technological CAD, and then which is done by the simulations. Simulation can predict I-V current voltage and capacitant voltage curve in extremely short devices. You can make transistor maybe one day. If you're making a nanotransistor in [inaudible] , they may be costing huge money and one year. However, this TCAD device can predict the nanotransistor within a couple of weeks. Then based on your simulation, you can modify the device structure before fabrication. Amazingly, modern TCAD technology is pretty accurate, and you can reduce the huge trial error of the future fabrications.
