Squares may be unfashionable, but for electronic circuitry no other shape will do. A method for making square arrays of polymeric nanoparticles could herald the next generation of miniature silicon chips.
Just ten years ago, who would have thought that you would be able to carry around your entire music collection in your shirt pocket, or your complete genome on a key ring? These amazing feats depend on miniature information-storage devices, which in turn are based on tiny integrated circuits. Although current technology is incapable of scaling down circuits much further, self-assembling nanoparticles could be pressed into service to make much smaller circuitry elements. But there's a problem — nanoparticles self-assemble into hexagonal arrays that are incompatible with the square arrangements used in industry-standard circuits. Reporting in Science, Tang et al.1 describe a method in which appropriately designed polymers form nanoparticles that assemble into square arrays.
The tiny structures found in integrated circuits are currently made by photolithography — a technique that uses ultraviolet light to create patterned masks from light-sensitive polymers, which are then used to control the etching of structures into the surfaces of silicon chips. The resolution of this technique is limited to about 100 nanometres (the wavelength of the light used to make the masks). But round nanoparticles can self-assemble into smaller patterns, packing together like oranges in a box. If this process could be used to make smaller features on circuits, it would make a huge difference to the amount of data that can be stored on a chip: a tenfold decrease in feature size could increase information density 1,000-fold.
Self-assembled structures are well known in nature — molecules such as lipids form both nanostructures (vesicles and micelles) and thin films (membranes). The synthetic analogues of lipids are block copolymers. Eleven years ago, it was shown2 that cheap processes can be devised in which vast arrays of these synthetic polymeric nanoparticles form spontaneously, creating structures ten times smaller than can be made by expensive photolithography. The electronics industry immediately saw the potential benefits of this, but was unable to find a cost-effective way of reworking circuit-design software and fabrication protocols to cope with the hexagonal arrays formed by self-assembly. The Semiconductor Industry Association therefore challenged researchers to develop a method for making square arrays of self-assembled block-copolymer particles. Tang et al.1 have risen to this challenge by using a combination of theoretical predictions and methods for controlling non-covalent interactions between molecules.
Like oil and water, most blends of two different polymers split into separate phases because of nonspecific, dispersive interactions between the two types of molecule. But by linking pairs of different molecules (A and B) with a covalent bond (to form an A–B diblock copolymer), one can prevent phase separation into two layers and gain some control over the nanostructures formed as the molecules aggregate. The nanostructures formed by such copolymers are the result of competition between molecular chain-stretching and interfacial effects around the covalent bond connecting the two polymer domains.
Much is known3 about the phase behaviour of diblock copolymers. In bulk material, the conditions needed to form three-dimensional cubic structures with long-range order are well established. In thin films deposited on substrates, where the thickness of the film is similar to the spacing between nanoparticles in the arrays, the preferred patterns are: hexagonal arrays of spheres; stripes of cylinders of block A embedded in a matrix of block B (with the stripes laid parallel to the film's surface); or single bilayers (or trilayers) of nanoparticles. The patterns observed depend on the composition of the copolymers and the interactions of the polymer molecules with the substrate and with the air above the film3.
But more complex molecules, such as A–B–C triblock copolymers — which contain three types of polymer-chain linked in series — can adopt a host of geometries inaccessible to simple A–B diblocks, including chiral and tetragonal arrays formed from cylindrical nanoparticles. Most importantly for forming patterned masks on circuits, A–B–C copolymers form stable, tetragonal arrays of cylinders in thin films. Moreover, the cylinders align perpendicular to the film's surface, unlike A–B diblocks, in which cylinders normally lie parallel to the surface.
There are, however, several problems associated with A–B–C triblocks. Not only are they more difficult to synthesize, but they also suffer from greater packing 'frustration' than do simple A–B diblocks, because the different blocks in each molecule prefer to pack together with blocks of the same type from other molecules. Domains of each block-type thus form, which means that each molecule must pass through two interfaces between different domains.
So although certain triblock compositions can make square arrays in thin films, long-range order is suppressed. Many of these issues can be resolved by making polymer blends of different diblock copolymers (A–B/B–C or A–B/C–D), which provide the advantages of having more than two distinct polymer blocks while avoiding the problems associated with connecting all the blocks in one molecule. But it has proved impossible to achieve long-range order in such blends because of their overwhelming propensity to generate separate phases, just as homopolymers do.
Tang et al.1 overcome this problem by using hydrogen bonding between a pair of A–B/B′–C diblocks. In their blend, blocks A and C are chemically different, and both are mutually incompatible with the B and B′ blocks. The authors' neat trick is to introduce small amounts of complementary hydrogen-bonding groups into the B and B′ blocks (Fig. 1a), so that the blocks form hydrogen bonds to each other. The resulting complex of diblocks behaves like a triblock, and phase separation of the diblocks is suppressed because the hydrogen bonding overcomes the nonspecific dispersive interactions.
The authors made their diblock copolymers using state-of-the-art methods to accurately control both the molecular weight of each block and the number of incorporated hydrogen-bonding groups. They then prepared the A–B/B′–C blend simply by mixing solutions of the two diblocks, and created thin films of the blend (about 50 nanometres thick) on silicon wafers. Using atomic force microscopy (AFM) to observe the nanoscale structure of the films, the authors showed that cylindrical nanoparticles in the film formed perpendicular, square arrays with long-range order (over an area of about 5 × 5 micrometres; Fig. 1b).
They obtained the best results when nearly equal numbers of complementary hydrogen-bonding units were incorporated into the B and B′ blocks. Blends that lacked complementary hydrogen-bonding groups formed only small nanostructured regions, which had hexagonal local order. The lack of nanostructure formation in such blends is probably caused by phase separation of the diblocks.
Tang and colleagues' numerical simulations1 of the packing in their blend reveal that square-packed cylinders have a lower free energy than hexagonally packed cylinders, unlike pure A–B diblocks. This is because a square lattice allows a more uniform distribution of C blocks around each A block (and vice versa), which minimizes unfavourable stretching of the B blocks even though the cylinders are not packed as closely as they could be. Although the A and C units are not connected on the same chain, dynamic hydrogen bonding ensures that, on average, an A–B diblock is always connected to at least one B′–C diblock. This favours B–B′ mixing at uniform chain-stretching, so that the blend mimics the behaviour of A–B–C triblocks.
The authors used their block-copolymer films as lithographic masks to transfer a template image into a silicon substrate. Using scanning electron microscopy, they observed that cylindrical pores (with diameters of about 22 nanometres) were formed in the substrate, spaced 50 nanometres apart. This was consistent with the patterns seen by AFM in the freshly formed films, and demonstrated a high degree of fidelity in the pattern-transfer process.
Could Tang and colleagues' approach be used in the mass production of integrated circuits? IBM already use block copolymers as masks for photolithography to make hexagonal arrays of cylindrical pores in insulators, so it seems that industry could easily adopt the authors' method. But one inherent problem is that of residual structural defects, especially when using large films — the larger the area covered by a mask, the more likely it is that a defect will occur on the underlying wafer. One solution could be to make mosaics of films from submillimetre-sized patches of polymers, so that the discovery of a defect doesn't require the whole wafer to be rejected; instead, any device using the wafer can be programmed to avoid using the defective 'tile'.
A combination of Tang and colleagues' technique with the latest technologies for making patterned diblock copolymers on patches and in channels could be used to make arrays of magnets that store trillions of bits of information per square centimetre. But this would be only the start: further developments will undoubtedly lead to greater miniaturization of silicon chips, and to the creation of electronic devices powerful beyond our imagination.
Tang, C., Lennon, E. M., Fredrickson, G. H., Kramer, E. J. & Hawker, C. J. Science 322, 429–432 (2008).
Park, M., Harrison, C., Chaikin, P. M., Register, R. A. & Adamson, D. H. Science 276, 1401–1404 (1997).
Russell, T. Curr. Opin. Colloid Interface Sci. 1, 107–115 (1996).
About this article
Journal of Heat Transfer (2018)
RSC Advances (2016)
Chinese Physics Letters (2014)
Journal of Polymer Science Part A: Polymer Chemistry (2013)
Polymer Nanowire Arrays With High Thermal Conductivity and Superhydrophobicity Fabricated by a Nano-Molding Technique
Heat Transfer Engineering (2013)