Paving the Way for Moore’s Law

Graphene

Carbon in the form of atom-thick “chicken-wire” sheets called graphene has the capacity to replace silicon circuitry.

As long as Moore’s Law holds true, every two years, computers will grow either twice as powerful or half the size. This trend, in which the number of transistors that can fit on an integrated circuit doubles every two years, has continued since the 1950’s and is forecast to continue for another decade. However, with the limits of silicon circuitry rapidly approaching the limits of manufacturing, some new material needs to take the place of silicon in order for Moore’s Law to continue. Carbon in the form of atom-thick “chicken-wire” sheets called graphene has the capacity to fulfill this role. It conducts electricity with very little resistance or heat generation at room temperature, consuming less power than silicon while allowing for high throughput. Though graphene-based circuitry is in its infancy, it holds tremendous promise as manufacturing processes continuously improve and new techniques emerge.

The limitations of silicon have been studied extensively, especially since the turn of the new millennium when silicon chips started to encounter performance issues. Silicon is used in transistors, which act like switches in circuit-boards, changing the way current flows by opening or closing a gate through an applied electric field. As the size of the silicon transistor shrinks, the resistance to current flow increases, so more power is consumed and more heat is generated. The major contributors to this effect are tunneling currents, which can pass through the gate even if it’s closed, while the increased heat promotes thermally generated sub-threshold currents, which leak in a similar way [1]. Until recently, fabrication methods have been changed and the size of the silicon elements could shrink without reducing their overall performance. However, silicon’s limits are becoming more apparent as everyday computing technologies progress. In order to develop a very small integrated circuit-board that is also powerful enough to do computing tasks quickly, like something found in the latest smart-phones, it will be necessary to develop some way to dissipate the extra heat, and to have an adequate, portable power supply to deal with the effects of shrinking the chip.

Graphene seems to be a viable successor to silicon because it can overcome the challenges that silicon faces as computer chips shrink. It was first “discovered” in 2004 in the form of “exfoliated” flakes that could be derived from graphite. It was well-known that graphite is a form of carbon where all the atoms are arranged in sheets and all the layers are loosely interconnected, but it was believed that none of these layers could be chemically separated without them decomposing into smaller fragments of carbon. For this reason, the carbon-sheet model was kept as a toy model for materials physicists [2]. However, when it was discovered that the layers of graphite could be successfully separated micro-mechanically while remaining stable in the form of atom-thick sheets [2], interest in graphene rekindled and soon roared to life.

The interest in graphene stemmed from its unique electronic properties. One of the most striking properties is that current is conducted through the monolayer with remarkably little resistance, even at room temperature. This arises from the way that electrons behave within the hexagonal grid that makes up graphene. When modeled, the behavior of the electrons more closely resembles particles called massless Dirac fermions, which can be thought of as electrons that lost their resting mass, or as neutrinos that gained an electron’s charge [2]. This behavior leads to virtually resistance-free flow of charge through graphene, making it very appealing for the manufacture of electronic devices. Moreover, the way that graphene conducts electricity depends on how large the sheet or ribbon of graphene is: if the ribbon is wider than a micron, it acts as a good conductor, and if the ribbon is thinner, it has the properties of a semi-conductor like silicon [3]. This is useful for potential electronics because it reduces the pileup of electrons that occur at the junction between the conducting and semiconducting material in traditional circuit boards [3], further decreasing the resistance in the chip as a whole.

Another property that makes graphene very promising for use in transistors is its capacity for ballistic computing. Ballistic computing puts forward a new design for transistors that allow them to function more rapidly and effectively and with fewer “leaking” currents than traditional transistors.

This revolves around a new design for the transistor gates themselves. The traditional design for transistors used in integrated circuitry is basically an electron basin with a metal gate: halting the flow fills the basin, designating a 1, while allowing charge to flow empties the basin, signaling 0. This introduces a delay to each operation, as the basin needs time to fill and empty to transmit the appropriate signal. Ballistic transistors offer an increase in the speed and efficiency by removing the need for halting the flow of electrons, instead using their inertia for “free” sorting into 0’s and 1’s [4]. The mechanism involves placing a wedge into the flow of electrons that would separate the flow into one of two directions, each designating either “1” or “0.” The direction of the flow is determined by an electric field upstream from the wedge that pushes the electrons slightly to one side or another [4]. The benefit of this setup is that very little energy is actually required to divert the flow from 1 to 0 or vice-versa, and since the flow is never halted, it can work at terahertz frequencies, which current transistors struggle to achieve [4]. While silicon sheets could work in this application, graphene’s pure crystal structure and high current capacity (while maintaining favorable electronic effects at room temperature) make it a better option [2].

The biggest hurdle for graphene electronics is adapting it to industrial processes. The exfoliated graphene flakes are irregular in shape and have rough edges. These are acceptable for research purposes, but regular and easily replicable shapes are necessary for industrialized processes because rough edges have a tendency to scatter the flow of electrons, introducing additional resistance to the system [2].

Recent developments have led to new ways of making variable sizes and shapes of graphene, which can be scaled to industrial proportions. The Golovchenko group at Harvard recently published a method that applied vaporized carbon to a nickel substrate where it forms a layer of variable thickness depending on the duration the vapor is present. Once the nickel cools, the carbon does not adhere to the crystalline structure and sizeable sheets of a uniform, determinate thickness are produced [5]. Walt de Heer’s group at Georgia Tech came up with a similar process, using etched silicon as the substrate. The carbon adheres to the surfaces etched into the silicon to generate graphene ribbons of a specific size with smooth edges, without any need for cutting procedures like the other methods to-date [3]. The use of silicon as a substrate also makes the process somewhat friendlier to industrialization, since most factories already have machinery for processing silicon.

Graphene holds a lot of promise for the future of electronics. Its electronic properties make it an ideal replacement for silicon as it reaches the limits of production. Graphene’s implementation in traditional and experimental electronic devices will result in electronics with smaller sizes and greater computing power. However, silicon will always remain an inexpensive and reliable material for circuitry, whereas graphene currently stands as a relatively expensive high-performance material. Eventually the industry will shift towards an all-graphene production as its price and ease-of manufacture approaches that of silicon, but this change will require a dramatic paradigm shift because the two technologies cannot be combined effectively. However, by the time Moore’s Law drives silicon to its physical limits, graphene circuitry will ideally be mature enough to extend the boundaries of computing possibilities.

References

  1. Wong PH. Device Scaling Limits of Si MOSFETs and Their Application Dependencies. Proceedings of the IEEE. 2001; 89(3):259-88.
  2. Geim AK, Novoselov KS. The Rise of Graphene. Nature Materials. 2007; 6:183-99.
  3. De Heer WA.  Scalable Templated Growth of Graphene Nanoribbons on SiC. Nature Nanotechnology. 2010; 5:727-31.
  4. Sherwood J. Radical ‘Ballistic Computing’ Chip Bounces Electrons Like Billiards. Univ. of Rochester News. 2006 Aug 16. Available at: http://www.rochester.edu/news/show.php?id=2585.
  5. Golovchenko JA, Hubbard W, Garaj S. Graphene Synthesis by Ion Implantation. Applied Physics Letters. 2010; 97:183-5.
  6. Image: AlexanderAIUS. Graphen. Wikimedia Commons; [uploaded 2010 Aug 26; cited 2011 Apr 28] Available at: http://commons.wikimedia.org/wiki/File:Graphen.jpg [Licensed under CC BY-SA]

Aleks Penev is a second-year biology and computer science major at the University of Chicago.