Fractal Nanotechnology
 GF Cerofolini^{1},
 D Narducci^{1},
 P Amato^{2} and
 E Romano^{1}Email author
Received: 24 July 2008
Accepted: 4 September 2008
Published: 23 September 2008
Abstract
Selfsimilar patterns are frequently observed in Nature. Their reproduction is possible on a length scale 10^{2}–10^{5} nm with lithographic methods, but seems impossible on the nanometer length scale. It is shown that this goal may be achieved via a multiplicative variant of the multispacer patterning technology, in this way permitting the controlled preparation of fractal surfaces.
Keywords
Introduction
Nanotechnology is essentially related to the problems of (1) predicting the properties of matter on the nanometer length scale, (2) preparing samples with accurately controlled size on that length scale, and (3) making accessible such samples to the macroscopic world. The first problem is essentially related to the fact that the nanoscale is characterized by the so called (N + 1) problem: the properties of a system with N particles may be largely different from those of a system with (N + 1) particles [1]. The preparation of nanoscopic samples with assigned properties requires therefore an extreme accuracy in the preparation.
The master road for the reproducible preparation of bodies in planar arrangement with assigned shape is photolithography. This technique has been able to produce features with progressively scaled size, the currently producible feature size being of several tens of nanometers. This size reduction, however, has been possible only thanks to the development of apparatuses of either huge economic cost (extreme ultraviolet lithography) or very low throughput (electron beam lithography).
In recent years, however, techniques not involving the use of advanced lithography have been developed for the preparation of nanometersized features. The most advanced ones are based on the transformation of vertical features into horizontal features and allow the preparation of lines with controlled width of 10–20 nm. Although this strategy allows the preparation of simple geometries only (line arrays), the development of new architectures [2] makes up for this inherent limitation. Among them the crossbar structure is particularly attractive [3–5] because it may simply be produced by crossing two perpendicularly oriented wire arrays and each crosspoint may be functionalized with the insertion of suitable molecules [6].
The first method for the nonlithographic preparation of ultradense line arrays was originally proposed by Natelson et al. [7]. It is essentially based on the sequential alternate deposition of two films A and B characterized by the existence of a preferential etch for one of them (say, A). After cutting at 90°, polishing, and controlled etching of A, one eventually gets a mold that can be used as a mask for imprint lithography (IL). Actually IL is a contact (rather than proximity) lithography; what is nonlithographic is uniquely the way used for the preparation of the mask. The first practical application of this idea was provided by Melosh et al. [8] who prepared a contact mask for IL with pitch of 16 nm by growing on a substrate a quantum well via molecular beam epitaxy, cutting the sample perpendicularly to the surface, polishing the newly exposed surface, and etching selectively the different strata of the well.
Another route for the nonlithographic preparation of ultradense line arrays is the multispacer patterning technique (S^{n}PT). This technology was developed by Cerofolini et al. [9, 10] with the goal of producing wire arrays with pitch on the nanometer length scale exploiting the already existing CMOS (complementary metaloxidesemiconductor) IC (integrated circuit) technology only. In this way the S^{n}PT may be viewed as a conservative extension of the current IC technology to the nanoscale length scale [11]. In the original formulation of S^{n}PT (hereinafter referred to as ‘additive route’, S^{n}PT_{+}) an array of 2n bars is directly defined onto a substrate via a sequence of n conformal depositions and anisotropic etchings. This idea is just an extension of the spacer patterning technique (SPT), conventionally used in microelectronics for the selfalignment of the gate electrode on sourceanddrain regions [12]. The density limit of crossbars prepared using S^{n}PT_{+} are discussed in Ref. [13]: although crosspoint density of 8 × 10^{10} could be achieved within the current technology, the overall process would however require 20 repetitions of each SPT cycle.
Multiplicative Route
Managing so many depositionetching cycles may however be difficult and expensive. Observing that the SPT allows, starting from one seed, the preparation oftwo spacers, the above difficulty may be removed using amultiplicative variant (referred to as S^{n}PT_{×}) of the multispacer patterning technique.
 (i)
the conformal deposition of a film on an assigned seed of high aspect ratio,
 (ii)
anisotropic etching of the film down to the appearance of the original seed, and
 (iii)
the selective etching of the seed.
with τ being the height loss per SPT_{×} cycle;τ depends on how accurately the technology has been set.
The first demonstrators of S^{n}PT_{×} for the generation of gratings sublithographic pitch go back to more than a score of years [14]; the usefulness of this technique for the preparation of wire arrays potentially useful for biochips, instead, is much more recent [15].
While the repetition in additive way ofn SPTs per (bottom and top) layers magnifies the lithographically achievable crosspoint density by a factor of (2n)^{2}, the repetition in a multiplicative way gives a magnification of 2^{2n }.
This matter is discussed thoroughly in another paper [13]; rather, in this work we intend to discuss another property of S^{n}PT_{×}—the possibility of preparing selfsimilar features with sublithographic definition.
The Multiplicative Route as a Technique for the Generation of Fractal Structures
Imagine for a moment that, in spite of the atomistic nature of matter and of the inherent technological difficulties, the multiplicative route can be repeated indefinitely. Remembering that the (n + 1)th step generates a set that is nothing but the one at the nth, , at a lower scale, the sequence defines a fractal; it will be referred to as multispace set fractal. Assuming the scaling law above [Eq. (3)], this fractal is selfsimilar only if the height of each spacer varies with n as 2^{−n }. Otherwise, if the structure scales only in one dimension or if its height scales with different law than in (Eq. (3)), the fractal is selfaffine [16]. As mentioned above, the ‘spontaneous’ height decreases with n (Eq. (1)) renders the fractal selfaffine. A selfsimilar fractal can be obtained at the end of process planarizing the whole structure with a resist and sputter etching in a nonselective way the composite film until the thickness is reduced to t _{0}/2^{ n }.
It is however noted that even ignoring the technological factors, the atomistic structure of matter limits the above considerations to an interval of 1–2 orders of magnitude, ranging from few atomic layers to the lower limits of standard lithography.
At each step the multispacer fractal set is characterized by a more uniform distribution of single intervals than the Cantor set; this makes the former more interesting for potential applications than the latter. In spite of that, trying to reproduce the Cantor set on the nanometer length scale seems of a certain interest. This is possible with existing technologies; as shown in Fig. 3, the process involves
(C1) the lithographic definition of seed (formed, for instance, by polysilicon) generating the Cantor set,
(C2) its planarization (for instance, via the deposition of a low viscosity glass and its reflow upon heating),
(C3) the etching of this film to a thickness controlled by the exposure of the original seed,
(C4) the selective etching of the original film,
(C5) the conformal deposition of a film of the same material as the original seed (polysilicon, in the considered example) and of thickness equal to 1/3 of its width,
(C6) its anisotropic etching, and
(C7) the selective etching of the space seed (glass, in the considered example).
Although the preparation of fractal structures may appear at a first sight nothing but a mere exercise of technology stressing, in the following we discuss some of their possible applications: (1) Systems biology, i.e. the study of the complex interactions in biological systems, is now overcoming the limits of molecular biology. Systems biology requires the knowledge of the cellular state at subcellular level (single constituting organelles: mitochondria, ribosomes, etc.). Sensing at this level, with space resolution in the deep submicrometer region, is impossible with CMOS devices, defined lithographically. Several attempts to overcome the CMOS limits are known [17, 18]; fractal technology, being able to transform a smart lithographic pattern (i.e. designed to the wanted function) in itself at a much lower scale, seems suitable for such a purpose. In particular, a matrix formed by the Cartesian product of two Cantor sets [16] would have nexttonothing contact area, thus providing sensing with minimum perturbation. (2) Superhydrophobic surfaces may be prepared controlling roughness and surface tension of nonwetting surfaces [19–21]. Whereas surface tension is a material property, roughness can be controlled by the preparation. For instance, roughness may be imparted depositing a suitable relief on the surface. In this way, however, an amount of area is lost for other application. Although it is possible that the control of wetting properties does not require to manage geometries on the nanoscale, this loss may be minimized designing the relief in such a way as to have almost nil area (as in the example above). (3) If the S^{ n }PT is used for the preparation of crossbar structures for molecular electronics, the functionalization with organic molecules of the crosspoints can only be done after the preparation of the hosting structure. According to the analysis of Ref. [22], this requires an accurate control of the rheological and diffusion properties in a medium embedded in a domain of complex geometry. Understanding how such properties change when the size is scaled and clarifying to which extent the domain can indeed be viewed as a fractal (so allowing the analysis on fractals [23] to be used for their description) may be a key point for the actual exploitation of already producible nanometersized wire arrays in molecular electronics.
Discussion
 (a)
a 2 × 2 crossbar obtained by crossing lithographically defined lines;
 (b)
a 16 × 16 crossbar obtained via S^{8}PT_{+}starting from lithographically defined seeds separated by a distance allowing the optimal arrangement of the wire arrays;
 (c)
a 16 × 16 crossbar obtained via S^{3}PT_{×}starting from lithographically defined seeds separated by a distance satisfying Eq. (2), and
 (d)
a 16 × 16 crossbar obtained via S^{3}PT_{×}starting from lithographically defined seeds and arranging the process to generate the Cantor middlethird excluded set.

the lithographic lines in (a) and (b) have width at the current limit for largevolume production, sayW= 65 nm;

the height loss τ is such that the maximum number of repetitions in the additive route is 8, and the sublithographic pitch is the same as reported in Refs. [9, 10]; and

the lithographic width of (c) is chosen to allow the minimum pitch to be consistent with the one obtained with the additive route (W = 100 nm), in this way producing sublithographic wires with width (12.5 nm) that has been proved to be producible [15].
Figure 4shows that the multiplicative route succeeds in producing crossbars with crosspoint density comparable with that achieved with the additive route, however, using a remarkable smaller number of SPT repetitions. To estimate the advantage of S^{ n }PT_{×}over S^{ n }PT_{+}, consider for instance the case of the repetition of three SPT_{×}per layer. This would produce a magnification of the lithographic crosspoint density by a factor of 2^{3} × 2^{3}. TakingW= 0.1 μm, after 3 SPT_{×}repetitions the spacer width should be of 12.5 nm, with minimum separation of 25 nm. Taking into account Eq. (2), the crosspoint density achievable with the repetition of 2 × 3 SPT_{×}would thus be nearly the same as that obtainable with the repetition of 2 × 8 SPT_{+}.
Conclusions
Some fractals (Cantor set, Peano and Koch curves, etc.) were known in mathematics well before the construction of a comprehensive theory [16]. Actually, the theory of fractals affirmed as such that only after Mandelbrot observation many physical phenomena can be described, although approximately and on a limited length scale, as fractals [24].
The usefulness of fractal algorithm is well known: assuming that human hairpins can be described, at least approximately as fractals, the use of fractal generator, rather than of the whole image, would greatly simplify storage and transmission of the corresponding image.
That real systems may be pictured as fractal set on the nanometer (and thus microscopic) length scale was first demonstrated by the analysis by Avnir, Farin and Pfeifer for the surfaces of several porous adsorbents [25–27]. In this work we have shown that the multiplicative route of the multispacer patterning technique allows the preparation of ordered fractals on the mesoscopic length scale.
Although at this stage we have no ideas of the possible practical applications of fractal technology, we nonetheless believe that the possibility of preparing, without the use of advanced lithography, fractal structures at the mesoscopic scale opens a virgin field of applications. The above considerations are certainly highly speculative, but not so speculative as those contained in van Gulick’s paper in topochemistry [28], outlining applications currently not achievable at that time, but later demonstrated to be possible.
Authors’ Affiliations
References
 Stoneham AM: Mater. Sci. Eng. C. 2003, 23: 235. 10.1016/S09284931(02)002746View ArticleGoogle Scholar
 Forshaw M, Stadler R, Crawley D, Nikolic K: Nanotechnology. 2004, 23: S220. 10.1088/09574484/15/4/019View ArticleGoogle Scholar
 Heath JR, Kuekes PJ, Snider GS, Williams RS: Science. 1998, 280: 1716. COI number [1:CAS:528:DyaK1cXjslOit7k%3D] 10.1126/science.280.5370.1716View ArticleGoogle Scholar
 Likharev KK: Nano and Giga Challenges in Microelectronics. Edited by: Greer J., Korkin A., Labanowsky J.. Elsevier, Amsterdam; 2003.Google Scholar
 Kuekes PJ, Stewart DR, Williams RS: J. Appl. Phys.. 2005, 97: 034301. 10.1063/1.1823026View ArticleGoogle Scholar
 Cerofolini GF: Appl. Phys. A. 2007, 86: 31. COI number [1:CAS:528:DC%2BD28XhtF2nsr%2FL] 10.1007/s0033900637364View ArticleGoogle Scholar
 Natelson D, Willett RL, West KW, Pfeiffer LN: Appl. Phys. Lett.. 2000, 77: 1991. COI number [1:CAS:528:DC%2BD3cXms12ht78%3D] 10.1063/1.1312256View ArticleGoogle Scholar
 Melosh NA, Boukai A, Diana F, Gerardot B, Badolato A, Heath JR: Science. 2003, 300: 112. COI number [1:CAS:528:DC%2BD3sXis1Cgt78%3D] 10.1126/science.1081940View ArticleGoogle Scholar
 Cerofolini GF, Arena G, Camalleri M, Galati C, Reina S, Renna L, Mascolo D, Nosik V: Microelectr. Eng.. 2005, 81: 405. COI number [1:CAS:528:DC%2BD2MXntlens7s%3D] 10.1016/j.mee.2005.03.041View ArticleGoogle Scholar
 Cerofolini GF, Arena G, Camalleri M, Galati C, Reina S, Renna L, Mascolo D: Nanotechnology. 2005, 16: 1040. COI number [1:CAS:528:DC%2BD2MXhtVWgsLvJ] 10.1088/09574484/16/8/007View ArticleGoogle Scholar
 Cerofolini GF: Nanotechnol. ENewslett.. 2005, 7: 5.Google Scholar
 Hunter WR, Holloway TC, Chatterjee PK, Tasch AF: IEEE Electron Device Lett.. 1981, 2: 4. 10.1109/EDL.1981.25319View ArticleGoogle Scholar
 Cerofolini GF, Amato P, Romano E: Semicond. Sci. Technol.. 2008, 23: 075020. 10.1088/02681242/23/7/075020View ArticleGoogle Scholar
 Flanders DC, Efremow NN: J. Vac. Sci. Technol. B. 1983, 1: 1105. COI number [1:CAS:528:DyaL2cXpsVCgsA%3D%3D] 10.1116/1.582643View ArticleGoogle Scholar
 Choi YK, Lee JS, Zhu J, Somorjai GA, Lee LP, Bokor J: J. Vac. Sci. Technol.. 2003, 21: 2951. COI number [1:CAS:528:DC%2BD2cXptFOlug%3D%3D] 10.1116/1.1627805View ArticleGoogle Scholar
 Falconer K: Fractal Geometry: Mathematical Foundations and Applications. Wiley, New York; 2003.View ArticleGoogle Scholar
 Hood L, Heath JR, Phelps ME, Lin B: Science. 2004, 306: 640. COI number [1:CAS:528:DC%2BD2cXos1KqtLY%3D] 10.1126/science.1104635View ArticleGoogle Scholar
 Stern E, Klemic JF, Routenberg DA, Wyrenbak PN, TurnerEvans DB, Hamilton AD, LaVan FA, Fahmy TM, Reed MA: Nature. 2007, 445: 519. COI number [1:CAS:528:DC%2BD2sXhtFWhtb8%3D] 10.1038/nature05498View ArticleGoogle Scholar
 Wenzel RN: Ind. Eng. Chem.. 1936, 28: 988. COI number [1:CAS:528:DyaA28Xkslentg%3D%3D] 10.1021/ie50320a024View ArticleGoogle Scholar
 Cassie ABD, Baxter S: Trans. Faraday Soc.. 1944, 40: 546. COI number [1:CAS:528:DyaH2MXhsFKqsA%3D%3D] 10.1039/tf9444000546View ArticleGoogle Scholar
 Tuteja A, Choi W, Ma M, Mabry JM, Mazzella SA, Rutledge GC, McKinley GH, Cohen RE: Science. 2007, 318: 1618. COI number [1:CAS:528:DC%2BD2sXhtlyrsr3E] 10.1126/science.1148326View ArticleGoogle Scholar
 Cerofolini GF, Casuscelli V, Cimmino A, Di Matteo A, Di Palma V, Mascolo D, Romanelli E, Volpe MV, Romano E: Semicond. Sci. Technol. 2007, 22: 1053. COI number [1:CAS:528:DC%2BD2sXhtlars7fJ] 10.1088/02681242/22/9/013View ArticleGoogle Scholar
 Kigami J: Analysis on Fractals. Cambridge Univ. Press, Cambridge; 2001.View ArticleGoogle Scholar
 Mandelbrot BB: The Fractal Geometry of Nature. Freeman, New York; 1982.Google Scholar
 Pfeifer P, Avnir D: J. Chem. Phys.. 1983, 79: 3558. COI number [1:CAS:528:DyaL3sXls1Cntbo%3D] 10.1063/1.446210View ArticleGoogle Scholar
 Avnir D, Farin D, Pfeifer P: J. Chem. Phys.. 1983, 79: 3566. COI number [1:CAS:528:DyaL3sXls1Cntbs%3D] 10.1063/1.446211View ArticleGoogle Scholar
 Avnir D, Farin D, Pfeifer P: Nature. 1984, 308: 261. COI number [1:CAS:528:DyaL2cXhsVGmsL8%3D] 10.1038/308261a0View ArticleGoogle Scholar
 Van Gulick N: New J. Chem.. 1993, 17: 619.Google Scholar