Information Technologies
Abstract
Since the intrinsic lower limit of the representation of one bit of information is around the atomic scale, nanotechnology offers particularly clear advantages through the miniaturization of information technology hardware. This limit is far from being reached, but the miniaturization so far achieved (feature sizes as small as 16 nm) already creates a problem of heat management because of the very high density of integrated components on a chip. Nanotechnology can help to solve this problem by creating thermal interface nanomaterials with very high heat conductivity. In the longer term, a move to spintronics embodied in nanostructures will eliminate heat dissipation altogether. Nanotechnology is also deployed in devices transducing information encoded optically or chemically into electrical form.
Keywords
Nanification; Vastification; Spin transistors; Flexible electronics; Data storage; Quantum dots; Display; Internet of things
It has already been pointed out that in information processing (including data storage) applications, nanotechnology offers many advantages because the intrinsic lower limit of the representation of one bit of information is around the atomic (nano) scale. The process of nanification of information-processing technology is well represented by Moore's law—which in its original form states that the number of components (i.e., resistors, capacitors and resistors) per chip doubles each year (a manifestation of vastification) [1]. When Moore revisited this prediction 10 years later [2], he somewhat refined his statement (since about 1972 up to the present, the number of transistors per chip has roughly doubled every two years), pointing out that the result was the consequence of three technological developments: increasing area per chip, decreasing feature size (defined as the average of linewidth and spacewidth), and improved design of both the individual devices and the circuit. Only the second of these three is a nanification process.
The direct economic consequence of these technological developments is a roughly constant cost per area of processed silicon, while the processing power delivered by the chip becomes steadily greater. Furthermore, nanification makes the transistors not only smaller, but also lighter in weight, faster (because the electrons have less distance to travel), less power-hungry, and more reliable. These are all strong selling points. Therefore, although technology push is undoubtedly important in maintaining Moore's law, the ultimate driver is economic.
As a result of these developments the microprocessor, which nowadays contains nanoscale components [3], has become ubiquitous throughout the world, even to the extent that a small company employing fewer than 50 people probably uses a computer to administer salaries, etc., despite it being almost certainly cheaper and more effective to do it manually.
8.1 Silicon Microelectronics
The starting point of chip production is the so-called wafer, the circular disc cut from a single crystal of silicon with a diameter of at least 300 mm and a thickness typically between 500 and 800 μm. Using lithography and etching technology the structures of integrated circuits are built up layer by layer on the surface of a chip [4]. Transistor construction has been based on complementary metal–oxide–semiconductor (CMOS) technology for decades. The size of the smallest features [5] in production has been steadily diminishing: 65 nm in 2005, 45 nm in 2007, 32 nm in 2009, 22 nm in 2011, and 16 nm in 2014 (at Intel)—this last value is close to the operational limit for metal–oxide–semiconductor (MOS) field effect transistor (FET) technology, but smaller sizes (10 nm and 7 nm) are being worked on. These developments represent tremendous technological challenges, not only in the fabrication process itself, but also in testing the finished circuits and in heat management—a modern high-performance chip may well dissipate heat at a density of 100 W/cm2, greater than that of a domestic cooking plate.
Silicon itself is still foreseen as the primary semiconducting material (although germanium, gallium arsenide, etc., continue to be investigated), but in order to fabricate ever-smaller structures, new photoresists will have to be developed. Furthermore, the silicon oxide thin film, which insulates the gate from the channel in the field effect transistor, becomes less and less effective as it becomes thinner and thinner (of the order of 1 nm). Other metal oxides (e.g., hafnium oxide) are being investigated as alternative candidates.
Some design issues arising from this relentless miniaturization are discussed in Chapter 10.
8.2 Flexible Electronics
Although the commercial sheen of wearable devices seems now to be in decline [6], there is still strong interest in components and circuits being developed with the original motivation for use in flexible, hence wearable, electronics, above all because they offer the chance of mass production via printing at much lower cost than established semiconductor processing. Since it is highly convenient if the printing substrate can be rolled up and unrolled (flexography), flexibility becomes a desirable attribute for the manufacturing stage. Furthermore, given the scarcity of indium, used in the tin-doped indium oxide transparent glasses required for one electrode in display devices and photovoltaic cells, the alternative, of dispersing a small volume fraction of highly conductive nano-objects (such as silver nanowires [8]) in an otherwise insulating, but transparent, matrix, has been enthusiastically investigated. The matrix is most conveniently an organic polymer, where flexible ones are also successful from the viewpoint of their other attributes. Hence, flexibility is available even if not required. Printing technologies are continuously progressing with the help of nanotechnology components, such as carbon nanotubes [7].
8.3 Heat Management
Heat management has grown in importance with the nanification of electronics, because the concomitant increase in the number of components per unit chip area implies a parallel increase in the heat generated (by the passage of electron currents) per unit area.
Since the surface of the object emitting heat inevitably has a certain roughness, contact with any juxtaposed heat sink will be imperfect. As in tribology, only the contact asperity is relevant, and its area may be quite small. The purpose of thermal interface materials is to fill the gaps. These materials are typically pastes of highly conducting nano-objects (carbon nanotubes are especially useful—cf. Table 5.1)—dispersed in silicone grease. A 10% volume fraction of the filler can increase thermal conductivity of the material from about 0.2 W/mK to almost 2 W/mK. The use of thermal interface materials (TIM) can typically diminish the interfacial thermal resistance by an order of magnitude (around 5 mm2 K/W is achievable). TIMs are essentially a kind of nanofluid that is preferably solid and rigid when cold.
Carbon nanotube arrays exploit the extremely high thermal conductivity of carbon nanotubes. They are synthesized directly on the silicon or silica substrate and perpendicular to it, forming a very efficient bridge between the electronic component and the heat sink [9].
8.4 Data Storage Technologies
Electrons have spin as well as charge. This is of course the origin of ferromagnetism, and hence magnetic memories, but their miniaturization has been limited not by the ultimate size of a ferromagnetic domain but by the sensitivity of magnetic sensors. In other words, the main limitation has not been the ability to make very small storage cells, but the ability to detect very small magnetic fields.
The influence of spin on electron conductivity was invoked by Nevill Mott in 1936, but remained practically uninvestigated and unexploited until the discovery of giant magnetoresistance (GMR) in 1988. The main current application of spintronics (loosely defined as the technology of devices in which electron spin plays a rôle) is the development of ultrasensitive magnetic sensors for reading magnetic memories. Spin transistors, in which the barrier height is determined by controlling the nature of the electron spins moving across it, and devices in which logical states are represented by spin belong to the future (Chapter 18).
Giant magnetoresistance (GMR) is observed in thin (a few nanometers) alternating layers (superlattices) of ferromagnetic and nonmagnetic metals (e.g., iron and chromium) [10]. Depending on the width of the nonmagnetic spacer layer, there can be a ferromagnetic or antiferromagnetic interaction between the magnetic layers, and the antiferromagnetic state of the magnetic layers can be transformed into the ferromagnetic state by an external magnetic field. The spin-dependent scattering of the conduction electrons in the nonmagnetic layer is minimal, engendering a small resistance of the material, when the magnetic moments of the neighboring layers are aligned in parallel, whereas for the antiparallel alignment the resistance is high. The technology is nowadays used for the read–write heads in computer hard drives. The discovery of GMR depended on the development of methods for making high-quality ultrathin films (such as molecular beam epitaxy).
A second type of magnetic sensor is based on the magnetic tunnel junction (MTJ), in which a very thin dielectric layer separates ferromagnetic (electrode) layers, and electrons tunnel through the nonconducting barrier under the influence of an applied voltage. The tunnel conductivity depends on the relative orientation of the electrode magnetizations and the tunnel magnetoresistance (TMR): it is low for parallel alignment of electrode magnetization and high in the opposite case. The magnetic field sensitivity is even greater than for GMR. MTJ devices also have high impedance, enabling large signal outputs. In contrast with GMR devices, the electrodes are magnetically independent and can have different critical fields for changing the magnetic moment orientation. The first laboratory samples of MTJ structures (NiFe–Al2O3–Co) were demonstrated in 1995.
8.5 Display Technologies
The results of a computation must, usually, ultimately be displayed to the human user. Traditional cathode ray tubes have been largely displaced by the much more compact liquid crystal displays (despite their disadvantages of slow refresh rate, restricted viewing angle and the need for backlighting). The main contemporary rival of liquid crystal displays are organic light-emitting diodes (OLEDs). They are constituted from an emissive (electroluminescent), conducting, organic polymer layer placed between an anode and a cathode (see also Chapter 16).
Any light-emitting diode requires one of the two electrodes to be transparent. Traditionally tin-doped indium oxide (ITO) has been used, but the world supply of indium is severely limited, and at current rates of consumption may be completely exhausted within 2 or 3 years. Meanwhile, relentless onward miniaturization and integration make it more and more difficult to effectively recover indium from discarded components. Hence there is great interest in transparent polymers doped with a small volume percent of carbon nanotubes to make them electrically conducting (see Section 5.1.4).
A significant advance in liquid crystal display technology has been made by incorporating photoluminescent quantum dots (QDs) into them, eliminating the need for white light-emitting diode backlights and color filters; a range of differently sized QDs is illuminated by a blue LED backlight, which is much cheaper, and according to size the QD emits a pure color. The launch of devices incorporating such displays has been an immediate commercial success.
“Electronic paper” or “e-paper” is based on a reflective, rather than transparent, liquid-crystal display.
8.6 Molecule or Particle Sensing Technologies
Information technology has traditionally focused on arithmetical operations, but information transduction belongs equally well to the field. Information represented as the irradiance of a certain wavelength of light, or the bulk concentration of a certain chemical, can be converted (transduced) into an electrical signal. From careful consideration of the construction of sensors consisting of arrays of discrete sensing elements, it can be clearly deduced that atomically precise engineering will enable particle detection efficiency to approach its theoretical limit [11]. Since a major application of such sensors is to clinical testing, they have been dealt with in more detail in Chapter 6.
8.7 The Internet of Things
The Internet of things (Iot) and the industrial Internet of things means the interconnexion of devices (machines). Insofar as it requires enormous computing power to handle the Internet traffic generated, it may be considered as an indirect manifestation of nanotechnology. It should greatly facilitate the development of servitization, in which an equipment manufacturer no longer sells the equipment to a customer, but sells the attribute produced by the equipment, such as tractive effort, or propulsive force, or a cool environment. The Iot makes this much more commercially feasible by allowing the manufacturer to continuously monitor the state of the equipment, facilitating rapid intervention if anything goes wrong and the optimal scheduling of routine maintenance. As experience in analyzing the resulting large volumes of data accumulates, it may become possible to predict incipient malfunction or failure, and intervene before it actually happens.
In comparison, the nonindustrial or domestic Internet of things seems of trivial importance. Applications are limited to such things as a refrigerator that can sense when supplies of comestibles are running low and order fresh supplies via the Internet, the supplies being delivered to the house by a delivery man or woman, or a drone. Although this may suit people with a narrow range of tastes and firmly entrenched habits (and recent surveys reported in newspapers show that an astonishingly high proportion of adults (of the order of 50%) eats the same food every day), it is unlikely to offer any benefit to the rest of the population cognizant of the vast variety of available foodstuffs and the even greater variety of circumstances and emotions that influence our choice of what to eat. It is certainly conceivable that objects like the domestic refrigerator could be endowed with artificial intelligence to enable them to respond more adventurously to stimuli, but this would doubtless require months or even years of machine learning. Besides, the last hundred years or so has seen a considerable contraction of the variety of comestible plants, fruits, etc., available commercially and nowadays it may take considerable effort to source foods that have become scarce. If anything, the domestic Iot is likely to exacerbate the contraction of variety. The contraction is not inevitable but experience shows that it is highly likely.