Microelectronics technology
The first M. products were created on bipolar transistors, because the quality of the silicon-dielectric interface plays a paramount role in the operation of field-effect transistors with a metal-dielectric-semiconductor (MDS) structure, and it took a lot of efforts from scientists and engineers to create technology for silicon surface preparation and methods for producing a high-quality thin layer of silicon dioxide (SiO2) on its surface. In the end. In the late 1980s, TFT technology was becoming more common in microprocessor and memory digital ICs, although bipolar ICs continued to be used in analogue and high-speed digital devices. Both technologies have certain advantages. Bipolar ICs offer high speed performance with large capacitive loads, while TIR ICs, particularly CMDP ICs (complementary TIR, i.e., complementary p-MTIR and p-MTIR transistors in series in a power-potential point circuit) offer low power consumption in static mode. In view of this, another technology is being developed (so-called BiCMOS), which enables both bipolar transistors and metal oxide semiconductor transistors (MOSFETs) of both conductivity types on the same chip and combines advantages of both technologies, but which is of limited use due to high cost. The silicon-on-insulator (SNI) IC technology was developed to create radiation-resistant M. products, initially in silicon on sapphire, then in silicon with a hidden dielectric layer, usually SiO2. This technology is applicable to ICs with high radiation resistance as well as to ICs with MR of 32 nm or less as there is no alternative.
TIR IC technology is evolving in two directions. One of them is related to achieving extremely high speed performance, which is limited by the heat removal capabilities, because in CMOS circuits the power consumption in dynamic mode is proportional to the operating frequency. Therefore personal computer processors are limited to an operating frequency of the order of 1 GHz when air cooling is used, allowing power removal of up to 50-80 W per 1 cm2 of the chip. In supercomputers, water jet cooling methods allow to remove up to 400 W/cm2, and the operating (clock) frequency of ICs can be increased by 5-7 times. Increasing the performance of processor ICs is also possible by “architectural means”, e.g. by creating so-called multicore processors, in which parallel calculations are implemented for a certain class of tasks; such processors are increasingly used in computing technology.
The main purpose of another trend is achieving extremely low power consumption, which is dictated by relevant applications (for example, in on-board equipment). Continuous increase in the number of transistors on a chip (at channel length of 6-7 nm this number will reach 1 trillion) forces the developers of ICs for computing machinery to find solutions to reduce power consumption and provide high performance (up to 100 Gflop). Otherwise, every computer center with a supercomputer would require a nuclear power plant. One of the possible solutions to this problem is to create chips with very low power consumption by lowering the supply voltage (e.g. to 0.4V at a threshold voltage of 0.3V) and thus significantly reducing operating frequencies, but ensuring high performance at the expense of multi-core processor architecture (up to hundreds of cores each). Cores most responsible for the performance can be designed to be fast.
Reducing power consumption, and consequently the heat generated by the ICs, opens the way to new chip assembly technologies. In particular, three-dimensional chip assemblies (so-called 3D assemblies) are becoming possible, in which thinned chips are placed one above the other, and in the isolation layers between them the so-called global connections are made. This technology allows the division of an integrated system into chips and the optimization of the connections.
A special niche in M. technology is occupied by semiconductor memory. Rapid, permanent and reprogrammable (flash) memory can be distinguished between super operational (cache) memory, operational, permanent and reprogrammable (flash) memory. Cache memory is made in the form of fast static memory with random sampling. Memory elements (usually based on CMOS technology) must be permanently connected to a power supply to store information. Microprocessors have a relatively small on-board memory (8-32kbytes), although an additional memory capacity (approx. 1kbyte) can be installed on the PC motherboard. Additional controllers with higher capacity (about 1 Mbyte) can be installed on the motherboard. Permanent memory (PP) – non-volatile, read-only memory that stores unchanging data.