Reports of the death of silicon electronics may well have been exaggerated. A technique that allows the deposition of silicon films from solution could harbinger the era of the inkjet-printed circuit.
On page 783 of this issue, Shimoda et al.1 set forth a radical way of incorporating silicon into that most basic of electronic components, the transistor. Their technique uses a novel liquid precursor of solid silicon to allow the ‘printing’ of semiconductor films via familiar inkjet technology. It could thus permit unprecedented control over the size and placement of semiconducting silicon in future generations of high-performance electronic equipment.
When its structure is delicately disrupted, ultra-pure silicon is the quintessential semiconductor. As such, it controls the electrical impulses that in turn control the computers and other electronic devices that many of us take for granted. Semiconducting silicon is obtained from highly purified natural silicon — which occurs in the form of silica (silicon dioxide) in quartzite rock and sands — by adding tiny amounts of appropriate impurities (the process known as doping), or through specialized crystallization methods. Whichever way is chosen, enormous effort goes into extracting, refining, shaping and processing silicon to make it technologically useful.
Recently, the need for semiconducting transistors to help transform electricity into coloured light for displays and screens has begun to present a challenge for existing silicon manufacturing technologies. In particular, the demand for screens with ever-increasing pixel resolution that are thinner, brighter, wider and lighter — or even flexible — has stretched the solid-state patterning techniques used to produce silicon circuitry to the limit. It has also fuelled intense research into alternative, more processable semiconducting materials, including organic molecules and polymers2,3,4.
Electronic devices based on solid layers of silicon still provide the benchmark for semiconductor performance, however. Conventional techniques for their manufacture involve, for instance, heating ultra-pure silicon in a vacuum to create a mist of free silicon atoms that condenses onto a supporting surface, preferably an inexpensive plastic. But multiple refining and deposition steps are still just the start of the delicate and convoluted manufacturing process that leads to a finished transistor (see Fig. 4d on page 785). Once deposited, the solid film must be sliced or etched to produce circuit elements, and then attached to the rest of the electronic components. All this must be done without compromising silicon's semiconducting properties. That places severe limits on semiconductor thickness, patterning and connectivity, and therefore on advances in electronics design.
Controlling a liquid is generally easier than sculpting a solid. Sophisticated, high-resolution printing technologies already exist that could be used to introduce very thin layers of liquid semiconductor in complex patterns over a variety of surfaces. The advantage would be particularly great for the mass production both of very small devices and of displays that cover huge areas, as their transistor components must typically be no more than about a thousandth of a millimetre thick. In both such applications, it is difficult to consistently reproduce transistor size, thickness and pattern using solid-state techniques.
But how can silicon be produced in liquid form? Molten silicon is out of the question here: with its melting temperature of 1,414 °C, it would destroy the other components of the device, as well as the printing equipment. Shimoda and colleagues' solution1 is to delve into the realm of decades-old synthetic chemistry. They focus on a binary compound of silicon and hydrogen, Si5H10 or cyclopentasilane, that is liquid at room temperature. When baked at a temperature of 300 °C or higher, this compound loses hydrogen gas, leaving a residue of pure, elemental silicon. It is thus, apparently, an ideal liquid precursor for silicon thin films.
But there is a problem. Cyclopentasilane tends to boil off during baking, making it difficult to control the amount of silicon that is actually left behind. The authors circumvent this obstacle using a technique called ‘ring-opening’ polymerization chemistry. They shine light of ultraviolet wavelength on to the five-membered silicon rings in the cyclopenta-silane liquid, causing them to open and join end-to-end. The result is long, non-volatile chains, known as polysilanes, that have the characteristics of viscous oils or even solids. If this polymerization is halted part-way through, the polysilanes already produced dissolve in the remaining unconverted cyclopenta-silane, resulting in a solution from which an elemental silicon residue can form. The amorphous network of silicon atoms, a-Si, obtained does not have the optimum three-dimensional structure for semiconducting behaviour, and so, as a final step, high-intensity ultraviolet light is applied to rearrange it into a more ordered, polycrystalline form (poly-Si).
Shimoda and colleagues are thus the first to produce relatively high-performance silicon films by processing from solution. They first prepared films by simple spin coating — essentially, spraying a thin layer of solution onto a quartz surface before baking — and found that the properties of the films were comparable to those of high-quality poly-Si produced by conventional techniques. Although this performance was lower for films deposited using inkjet-printing technology, it was still much higher than is typically achieved for solution-processed films based on alternative, organic materials4.
This method dispenses with some high-temperature refining of metallurgical silicon extracted from silica and replaces it with chemical synthesis and milder distillations. Admittedly, this liquid polysilane precursor is highly sensitive to contamination by oxygen both during and after its preparation. Such contamination can drastically diminish the electronic performance of the eventual film, and dictates that air and water must be rigorously excluded at all stages of the process. These precautions are, however, no different from those taken for traditional routes to silicon thin films.
But it is the potential for taking advantage of highly controlled printing techniques in the patterning of silicon thin films that makes this work particularly exciting. Challenges do remain here: the physical properties of the liquid precursor, such as its viscosity, vapour pressure and surface tension, must be tuned to take full advantage of the printing tools available, while precisely controlling the post-baking thin-film morphology critical to semiconductor performance. Careful study of dilution effects is also needed. Finally, stricter control of the distribution of polysilane chain lengths within the precursor mixture would also probably help — a real test for polymer chemists.
It might be that, in the end, inkjet printing of ‘liquid silicon’ will not provide the resolution necessary to pattern a high-density integrated circuit and therefore make a computer chip. But what it will certainly allow is the remarkably straightforward generation of simple, cheap and flexible circuits for displays, as well as a range of other applications — solar cells, X-ray detectors and multi-analyte chemical sensors included.
Shimoda, T. et al. Nature 440, 783–786 (2006).
Horowitz, G. Adv. Mater. 10, 365–377 (1998).
Dimitrakopoulos, C. D. et al. Adv. Mater. 14, 99–117 (2002).
Mitzi, D. B. et al. Nature 428, 299–303
About this article
Thin Solid Films (2012)
Hydrogenated amorphous silicon nanostructures: novel structure–reactivity relationships for cyclization and ring opening in the gas phase
Theoretical Chemistry Accounts (2011)
Novel chemical precursors and novel CVD strategies enabling low temperature epitaxy of Si and Si:C alloys
Thin Solid Films (2010)
Journal of Propulsion and Power (2008)