Automation of integrated circuit design
As the complexity of silicon ICs increases, automation of IC design becomes increasingly important. In world practice, companies specializing exclusively in IC design and possessing advanced CAD systems are being created. These systems include transistor models (so-called SPICE-models), logic element libraries, the number of which can be up to 50-200, design and technology files to ensure compliance with topology standards, and libraries of complex functional blocks (so-called IP-blocks; IP – intellectual property), which allows to design ICs of any complexity, including SNC.
IP-blocks are ready-made models of devices that implement the functions of processors, memory modules, timers, parallel and sequential I/O ports, interface controllers, digital-to-analog devices, etc. Using the library set of IP-blocks, it is possible to design SNK, oriented on the effective performance of computational tasks and tasks of object control in real time. There is a continuous growth of sales volume in the world electronics industry in the form of SNK, which already by 2008 amounted to tens of billions of dollars. US.
Problems and prospects for further development. By 2020, reducing MMR in ICs will lead to an increase in the number of transistors up to 5-1010 in processor ICs and up to 1012-1013 in memory circuits. The channel length in such transistors will reach 6 nm. This raises serious problems. First, further reduction of the transistor channel length will lead to an exponential growth of the current between the transistor source and drain in the static state. This current is a direct tunneling current and can reach 10-6 A/µm channel width with a channel length of 5 nm, whereas the acceptable value of current between transistor source and drain in the off state is 10-9 A/µm. The result is an increase in static power consumption by almost 1 kW, which is released as joule heat and must be dissipated from the chip.
Secondly, the characteristics of a transistor with such a small size are affected by quantum effects. One such effect is the longitudinal quantum motion of the electron along the channel. The electron propagates in the channel as a de Broglie wave, which interferes and scatters on the roughness of the channel boundaries and on atoms of random impurities. The presence of random impurities and roughness, also having a random character, leads to scattering of transistor characteristics (the value of current in the open state), which, as the simulation shows, with only one impurity atom near the source can reach 20% (acceptable scatter is about 10%). This places very serious demands on the quality of the starting materials (Si in this case) and on the technology of their further processing.
It is likely that for MR less than 6-7 nm tunneling transistor with Schottky contacts on unalloyed silicon is satisfactory in its characteristics. On the basis of tunneling transistors, circuitry similar to CMOS is in principle possible.
There is an intensive search of other types of transistors. First of all they are devices based on graphene (a monolayer of graphite) and carbon nanotubes (rolled up into a tube graphene). The main motive for the use of these materials as transistor channels is the huge (about 105 cm2/V-c) mobility of charge carriers. However, so far it is not possible to create field-effect transistors with such materials as channels, capable of competing with silicon field-effect transistors. The possibility of using nanotubes, graphenes and the so-called graphene edges (narrow strips of graphene) as connections in ICs is also discussed.
Interest in the so-called one-electron transistors, which are based on one-electron tunneling (first described by K.K. Likharev) and the “Coulomb blockade” effect, is not diminishing. One-electron devices have a very high speed (in the limit it is the tunneling time of one electron) and very low power consumption (about 10-8 W). However, such devices operating at room temperatures have not yet been created due to the need for lithography with a spatial resolution of about 1 nm.
Molecular (including unimolecular) transistors are also under investigation, the “attractiveness” of which lies primarily in the complete identity of the size of molecules of a certain type (i.e. 100% size reproducibility). Molecular transistors, like transistors on carbon materials, may find application as highly sensitive sensors in the coming years.