Since the first integrated circuits were fabricated at Texas Instruments and Fairchild Semiconductor in the early 1960s,the number of ICs that can fit on a chip has doubled approximately every two years following the well-known Moore’s law. This has been made possible by reducing the minimum feature size. As of 2022, through advancements primarily related to the development of new processing techniques, the smallest transistors today are 2 nm – small enough to fit 50 billion on a chip the size of a fingernail. While ICs were first built using germanium, they were soon replaced by silicon for two key advantages. First, silicon is abundant in nature, providing the possibility of manufacturing electronics with a low-cost starting material. Second, the processing advantages of silicon raised it above other semiconducting materials. CMOS is the technology used to make most conventional electronics. The MOS and bipolar structures are fabricated through repeated application of several processing steps, such as Photolithography, Etching, Diffusion, Oxidation, Evaporation, Sputtering, Chemical Vapor Deposition , Ion Implantation, Epitaxial Growth, and Annealing. The strength of this manufacturing strategy lies in the standardization of these processes; It is possible to reliably and repeatably make micro- and even nano-scale features by carefully following a process recipe.
Despite the benefits,raspberry container there are some limitations to conventional electronic manufacturing technologies. For one, the processing temperatures to obtain critical components in these devices are hundreds to thousands of degrees Celsius, which limits material selection drastically. Of these limited materials, they are nearly all rigid at room temperature. Rigid materials are inherently difficult to integrate with biological applications; Nature is full of organic curves and fractal patterns, while conventional electronics are Cartesian and rectilinear. There are also practical limitations to the size of silicon-based electronics, leading to the largest commercially-available silicon wafer being only 450 mm in diameter. These primary limitations motivate printing technologies, which circumvent these problems and offer additional benefits.Because the processing temperatures are much less than those required for rigid silicon electronics, there are many more materials compatible with printing that would otherwise melt or incinerate at foundry temperatures. These materials can be solution-processed, i.e., the discrete material particles can be suspended or dissolved in a liquid-phase solvent and deposited onto a substrate – or ‘printed’. Printing is an additive process, meaning material is only deposited where it is used, and there are no steps that require removing material as in conventional lithographic processes. Many of these printing processes can easily be scaled to roll-to-roll processing, making it possible to produce large volumes of printed electronics at minimal cost with the ability to change the design of the printed device quickly.
Finally, the low temperatures of printed technologies also allow for using plastic, paper, or other flexible materials as the base substrate. Despite the benefits, there are very few fully-printed electronic systems in practice. It is challenging to build analogs to the transistor using printed technologies. While there has been a lot of research on developing printed transistors, more work needs to be done for wider adoption. Thus, most self-proclaimed printed electronic systems are hybrid electronic systems in disguise. A substrate is the material that is being printed onto. Substrates determine the bulk mechanical properties of the device, which is a large part of why flexible films are most commonly chosen. Polymer films, for example, can be flexible and/or conformal, which can be leveraged in the design of manufacturing processes and used in various application spaces where physical flexibility is essential. In other areas where flexibility is not necessary or specific substrate material is required for additional features in a device, it is still possible to print onto rigid substrates.When a surface’s surface energy is high, the printed ink will try to spread. When the surface energy is low, the ink will form islands or beads of ink. The material properties primarily determine the surface energy of a surface, though processing of the material can modify its surface energy.Conductors are the base structural block of all electronic devices – printed or otherwise. They carry the power that powers the device, form the interconnections between device layers, and transmit data in an electric signal. In printed electronics, the conductor is either made from nanocomposite ink or organic polymers. Nanocomposite inks are common conductors in printed electronics and are made up of conductive particles, polymer binders, a solvent, and sometimes other tuning components.
The suspended conductive particles, when printed, form a percolated network within the non-conductive polymer binder, while the solvent evaporates away. Silver, carbon, and copper inks are the most popular choices of conductive particles for conductors, though other metals are sometimes used as well. The selection of conductor and the particle:binder ratio can be altered to tune the conductivity of the printed feature. However, these changes affect other material properties of the composite as well.Dispenser printing encompasses all printing techniques that employ a semi-continuous flow of ink through a nozzle or print head. A schematic of the process is shown in 2.2F. The composition of inks in dispenser printers can vary widely, from solvent-less fused deposition modeling printers to binder-less direct ink writers , and everything in between. In any case, dispenser printing utilizes a nozzle mounted on a 2D- or 3D- chassis. Similar to inkjet printers, CAD is used to generate digital designs of the printed pattern, and the computer generates a program that controls the print head and the flow rate through the nozzle to create the desired pattern. An exciting application of dispenser printing is the incorporation of electroactive particles in the polymer filaments used in FDM-type 3D printers. In this method, the ‘ink’ is a solventless blend of 3D printing polymer and electroactive particles such as carbon black. The polymer pellets are heated beyond their melting point and mixed with the particles before being extruded into a wire-shaped filament that is coiled and later used in the FDM 3D printer. The 3D printing process then remelts the wire by Joule heating of the nozzle and mechanical pushing the filament through the nozzle at a controlled rate while the chassis moves the nozzle to the desired location of the pattern. DIW is a similar process to FDM printing with three distinct features. First, the ink of a DIW always includes a solvent to control the viscosity of the ink, though sometimes the solvent is a gel-phase material such as PVDF. Second, the material is physically pumped through the nozzle like squeezing ketchup through a bottle. Finally, the ink does not need to be heated in DIW, whereas the filament in FDM printing will only flow when brought past its glass transition temperature, and is generally brought close to or beyond the melting temperature of the polymer. These differences give some advantages to DIW. Because DIW can be performed at lower temperatures and the rheology can be tuned by changing the volume of solvent, a wider range of materials can be used. After printing, there are several heat-treatment steps that may be required to complete the printed component layer. The most common of these are annealing, curing, and sintering. Annealing is a heat-treatment process that is used to relieve the internal stresses of a material. Annealing is more commonly used for the heat-treatments of macro-scale metals,plastic plant pots ceramic glasses, and high-performance polymers, though it is also applied to printed electronic components as well.
A macro-scale example would be a cold-rolled steel billet annealed so that it can be worked further into final products. In printed electronics, annealing is more commonly used to reduce the internal stresses of the polymer binder or to reshape the crystallinity, such as PLA. Curing is a process that accelerates a chemical reaction, and in the case of printed electronics, it almost always refers to the cross linking of a thermoset polymer binder. Thermoset polymers that do not set in a reasonable time at room temperature are cured at higher temperatures in an oven or vacuum oven. The monomers react much more rapidly at the curing temperature, hardening it beyond what would otherwise be possible. Sintering is a process for causing nano- or micro-scale particles to become a monolithic bulk material by diffusion. Sintering may occur in either or both of the solid and liquid states. Consider the extreme case of a composite of perfectly sphere shaped particles in a polymer matrix. Regardless of the particle:polymer ratio, two perfect spheres can only contact one another at a single point. If these were the conductive particles in a printed conductor composite, then the resistance would be very high despite the inherent conductivity of the particles because the cross-sectional area would be infinitesimal. In sintering, the composite would be heated above the melting temperature of the conductive particles so that they diffuse into one another, increasing the cross-sectional surface area and improving particle-to-particle contact.Printing is a disruptive technology that is a complete change in how electronics are made. However, considerable advances in printed and hybrid electronics are still needed to deliver on its market promises.
Part of the problem is that printed electronics are always bench marked against conventional silicon microelectronics in areas that favor silicon. This is erred thinking; Printed electronics should not compete against conventional electronics in areas such as charge mobility for printed semiconductors, the conversion efficiency of printed solar cells, or minimum feature dimension size. While researchers should strive to improve these areas, the future of printed electronics lies in its advantages over conventional electronics: large area, flexible, and low-cost manufacturing of electronics with high volumes and a wider variety of viable materials. These advantages open a new world of possibilities that would be unpractical to achieve with silicon alone.Artificial Intelligence is the capability of a computer system to mimic human cognitive functions. Computer scientists commonly develop AIs to mimic how humans learn and solve problems. They do this by programming the computer system to use math and logic to simulate people’s reasoning to learn and make decisions. Machine Learning is a subcategory of AI. Specifically, it is the application of mathematical models to help a computer system improve – or ‘learn’- without direct instruction. This enables a computer system to continue improving on its own based on its previous results or experiences.AI is what an ‘intelligent’ computer system uses to behave and perform tasks like humans. ML is how a computer system builds its intelligence. There are three primary strategies for building an ML program. These are supervised learning, unsupervised learning, and reinforcement learning. Supervised learning has the defining characteristic of access to annotated training data. Supervised learning algorithms induce models from the training data, which can then be applied to classify other unlabelled test data. A common analogy for supervised learning is that of a teacher teaching a student: the teacher trains a student with lots of practice problems, and then the student takes a test without the teacher’s help. If supervised learning is analogous to a classroom learning environment, then unsupervised learning is like throwing a child into the deep end of the pool to teach them how to swim. Unsupervised learning modes do not have access to labeled training data. Instead, unsupervised learning algorithms learn by clustering the data together in different ways, and trying to find patterns. The child in the pool will learn that behavior where they tread their legs and wave their arms brings them closer to the surface. Finally, reinforcement learning is akin to training a dog with treats. A dog will act however it wants, but they will learn that certain desired behaviors will result in a treat, such as returning to their handler when they shout ‘come’ or reclining onto their haunches when they call ‘sit’. In a reinforcement learning program, the ‘treat’ is a numerical output. In practice, a reinforcement learning program will perform a task by trial-and-error, the task result will be scored, and then the program will attempt the same task again with behavior similar to its highest-scored behavior in previous trials. Regardless of the learning strategy, all machine learning algorithms follow a process flow similar to the one shown in Figure 3.1. A computer model is built and executed, the simulation results are scored by calculating an error term, and the error drives adjustment of the computer model parameters.