INTRODUCTION
Optical methods continue to surprise us. The Abbe limit, of approximately 1/2, was once thought to be an insurmountable barrier to their exploitation, but the ingenuity of physicists continues to confound this unduly pessimistic outlook. There are currently a number of tools, including near-field,[1] two-photon,[2] fluorescence-based,^ and plasmonic techniques[4] that all offer sub-diffraction limit spatial resolution. Moreover, while the end of photolithography has been forecast for about two decades, the semiconductor device industry remains firmly wedded to it as its primary manufacturing tool, and electronic engineers have been very creative in adapting it to facilitate the onward progress of their industry along the path charted by Moore’s Law. However, at the very smallest length scale, electron beam lithography has continued to be seen as the gold standard fabrication tool. While electron beam methods are serial ones and do not readily translate into a manufacturing environment, they have been thought to offer a degree of resolution beyond the reach of photolithography.
Here we review new work that suggests this view is incorrect. By using scanning near-field photolithography (SNP), it is possible to generate structures in organic monolayers that are comparably small to anything fabricated by electron beam lithography. Importantly, however, these approaches, based on the excitation of photochemical reactions, are capable of being implemented under ambient conditions or even under fluid. Although such near-field lithographic tools are in their infancy, they offer exciting prospects for use in lithography at extremely small length scales, along with a capability to manipulate biomolecular and organic molecular systems that are inaccessible to electron beam methods. Combined with multiplexing, they may, given their lower cost, offer much greater potential for commercial exploitation than has ever been the case for electron beam lithography, and they may prove a valuable tool in the drive toward highly miniaturized molecular devices and materials.
NEAR-FIELD SCANNING OPTICAL MICROSCOPY
In surface microscopy, near-field scanning optical microscopy (NSOM, also known as scanning near-field optical microscopy, or SNOM) has provided access to nanometer-scale optical characterization. The original concept for NSOM was developed by Synge, early in the 20th century. His ideas revolved around the use of small optical apertures to characterize materials. Ordinarily, when light is passed through an optical aperture smaller than the Abbe limit, it undergoes diffraction, imposing a lower limit on resolution (also called the diffraction limit). Synge proposed that an optical aperture could be used for subdiffraction limit imaging provided it was brought adequately close to a solid surface.[5-7] Under such conditions, illumination occurs in the near field, effectively meaning that light from the aperture interacts with the sample before diffraction can occur (Fig. 1). Although the underlying concept was very simple, its realization in practice was much more challenging, because for it to be effective, the separation between the aperture and the sample must be very small indeed (nm distances) and must be maintained within a very small range of tolerance during the course of the whole experiment. The empirical corroboration for Synge’s ideas was eventually provided in 1972 by Ash and Nichols,[8] who achieved a resolution of 1/60 with 3 cm microwaves. Although this result was impressive, it represented, at that wavelength, a resolution that was still a macroscopic distance.
A significant leap forward was made by Betzig and Trautman in 1992. They used an optical fiber, drawn to a narrow apex and coated with a metal film to constrain the electric field, to image samples.[9] They claimed that a resolution of 12 nm was feasible, but only after deconvolution of the probe geometry. This spatial resolution has unfortunately not been widely reproduced. While subdiffraction-limit imaging is feasible, and a variety of types of characterization are possible,[10] fiber-based approaches have generally proved difficult to implement. Generally speaking, the best resolution achieved has been approximately equal to the probe aperture, typically ca 50 nm, and usually, the resolution has been poorer than this—often significantly so. One of the problems of fiber-based methods is that the reproducible preparation of small apertures with high transmissions is difficult.
Recently, there has been a growing interest in the use of what are termed "apertureless” NSOM methods. These exploit the fact that irradiation of a metallic tip held in close proximity to a solid surface using a suitably polarized light may lead to a very pronounced enhancement of the electric field in a small region beneath the tip.[11] A variety of optical phenomena may be excited in this way. The tip may, for example, act as an antenna, a phenomenon referred to rather evocatively as the lightning-rod effect, yielding a strongly focused field. A surface plasmon mode may be excited at the tip. Again, the electric field associated with the plasmon excitation exhibits a pronounced confinement. Fluorescence resonance energy transfer between donor and acceptor systems attached to the tip and the sample surface may occur,[12] offering exciting prospects for the characterization of biological systems. Two-photon absorption[13] and other nonlinear phenomena, such as second harmonic generation,[14] have also been reported. While many of these techniques are currently difficult to implement, they offer extraordinary capability. For example, Hartschuh et al. carried out an apertureless Raman investigation of single-walled carbon nanotubes.[15,16] They reported Raman spectra from single nanotubes, with a spatial resolution of ca 25 nm. These results constitute an exciting advance for near-field microscopy and spectroscopy. Such progress suggests that after a long gestation period, NSOM may be on the verge of delivering the kind of fruit that it has long promised.
Fig. 1 By bringing a surface close to a small aperture, it is possible for light to emerge from the aperture without undergoing diffraction.
LITHOGRAPHY USING NSOM
In photolithography, light is typically directed through a mask at a layer of photoresist. The light exposes the resist, leading to either the cross-linking of oligomeric material or the removal of a photosensitive material, and the formation of a photopattern that may subsequently be used to transfer structures into the underlying substrate. Clearly, it is possible to fabricate smaller features by simply reducing the dimensions of the gaps in the mask through which the light passes, and for many years the semiconductor device manufacturers have been able to do this. However, eventually the diffraction limit is reached. At this point, light is diffracted as it passes through the mask. There are several solutions to this, including reducing the wavelength of the light used. In principle, a photolithography mask may be placed in contact with the sample (rather than being placed away from it, as in projection photolithography), to try to exploit near-field effects. However, the great problem with this is ensuring that the mask remains genuinely within the near-field regime across the whole of its area. The tiniest specs of dust or small variations in substrate topography would mean that significant portions of the mask were not in contact and diffraction would occur.
The adaptation of a scanning near-field optical microscope represents another solution. While mask-based processes offer the advantage of parallel fabrication of large numbers of structures, there are other applications for photolithography, outside of electronic device fabrication, where the ability to excite different processes at different locations on the sample is a distinct advantage (for example, the immobilization of biological molecules using light-directed processes).[17,18] Moreover, the development of high-speed scanning systems[19] and multiplexing technologies[20] both provide potentially important new routes to high-throughput fabrication.
In a fiber-based NSOM system, the aperture at the end of the optical fiber constitutes a nanoscopic light source. Provided the probe is maintained in close proximity to the sample at all times, the dimensions of the illuminated region will be determined, to a first approximation, by the dimensions of the probe. Given that the fabrication of probes with 50 nm apertures is technically feasible with reasonable repeatability, and that the transmission will in any case decrease sharply at smaller aperture dimensions, it would seem, a priori, feasible to aim for a resolution of ca 50 nm in well-optimized circumstances. The first report of the use of NSOM to conduct lithography, by Betzig et al.,[21] suggested that this was realistic. Using an optical fiber probe, controlled by shear-force modulation, they wrote structures into a Co/Pt multilayer film using visible light (488 and 514 nm). In regions heated near to the Curie temperature of the medium (ca 300° C), domains were formed with opposite magnetization that could subsequently be imaged using NSOM. When adequate powers (ca 5 mW input power from an argon ion laser) were utilized, it was possible to write features with diameters of 60 nm. Optimization of the conditions relied upon ensuring that adequate power was passed through the probe to ensure sample heating and, at the same time, that the input power was not great enough to damage the aluminum coating on the probe.
Building on this promising start, Krausch and coworkers utilized an NSOM to expose films of conventional photoresist.[22-24] Using an argon ion laser (1 = 454 nm), they were able to write structures as small as 1/5 (ca 80 nm) into Hoechst Novalack AZ 6612, a resist based on a phenol-formaldehyde resin, which becomes base-soluble on exposure to light. The features formed in the photoresist represented a mechanical replica of the intensity distribution in the optical near field of the tip. It was found that the features could be fitted with a Gaussian intensity distribution with a width of approximately 100 nm, and a height of ca 15 nm. A grating was fabricated with a period of 164 nm and a line width of 82 nm. These results confirmed that by using NSOM, it was possible to break the diffraction limit in lithographic mode by a significant margin.
Smolyaninov Mazzoni, and Davis utilized a standard negative-tone resin-based photoresist (Shipley KTI 747) in studies using UV light from an excimer laser (248nm).[25] Exposure of the resist using an NSOM led to the photochemical generation of a cross-linked network that was insoluble in the developer. The resulting specimens were characterized before development, by using shear-force imaging with the same probe employed to modify the sample, and after development, by AFM. A nonlinear dependence of the feature size on the light power was reported. Under optimal conditions, features of size ca 100 nm could be written with an uncoated optical fiber.
Although these results represented excellent progress, they nevertheless leave much to be desired. In particular, these early studies failed to realize the optimal resolution expected a priori—matching the diameter of the optical fiber aperture. One of the problems was that these photoresist films had a finite thickness—tens of nanometers at best. However, the electric field associated with an optical aperture in the near field is known to diverge comparatively rapidly within dielectric layers.[11] This may mean significant spreading of the excitation in the resist layer. Moreover, the approach of Betzig et al. relied upon localized heating; thermal migration away from the region exposed beneath the fiber aperture may lead to an additional broadening effect. These problems were realized by Fujihara and coworkers,[26,27] who sought to restrict the thickness of the resist by utilizing monolayer systems. They prepared Langmuir-Blodgett films of a photochromic material containing 4-octyl-4′-[5-carboxypentamethyleneoxy)azobenzene. On exposure to UV light (1 = 350 nm), this molecule undergoes a cis-trans conformational change. The change may be reversed by the action of heat or light. They reported a resolution of 130 nm. Although this does not represent an improvement on the work described above, it nevertheless suggests that such systems have a useful role to play.
Conjugated polymers have a broad range of potential applications arising from their electrical and optical properties. There has been significant interest in exploiting the susceptibility of conjugated polymers to photooxidation as a means for patterning them at the nanometer scale, using an NSOM to deliver excitation to localized regions in polymer films. Wei et al. studied thin films of (3,4-diphenyl-2,5-thienylene viny-lene). They reported comparatively diffuse structures with line widths of ca 200nm.[28] Buratto and coworkers patterned films of poly[2-methoxy, 5-(2'-ethyl hexyloxy)-p-phenylene vinylene] using light from an argon ion laser coupled to an NSOM.[29,30] They also reported a line width of 200 nm. In part, no doubt, the poor resolution in these studies was a consequence of the finite thickness of the resist. Credo et al.[30] also studied the exposure of tris-8-hydroxyquinoline aluminum (Alq3) films. Here there was an additional problem in that active species created by the initial exposure of the sample diffused away from the region of initial exposure. On small length scales, even small amounts of diffusion may cause a significant degradation of the resolution. In studies of self-assembled dye layers, they also reported degradation of resolution; but, in this case, it was because of the migration of energy through the resist material rather than the diffusion of active species. Once again, the finite film thickness (50 nm) may well account for much of the migration of heat energy. Studies of the near-field exposure of conjugated polymer films have generally failed to yield dimensions that are authentically nanometer scaled (i.e., less than 200 nm), and much of this can be attributed to the problems associated with the finite thicknesses of even the highest quality materials prepared by techniques such as spin-coating. More recently, for example, Riehn et al. patterned PPV layers with a resolution of, at best, 160nm.[31] They modeled the behavior using the Bethe-Bouwkamp model, and determined that a surface of constant intensity was formed, which just touched the substrate under a 40 nm precursor film and had an extent of approximately half the diameter of the feature prepared experimentally. They proposed the formation of a central core surrounded by a gel phase. The steep profile of the electric field predicted by their model indicated that the polymer at the top of the sample absorbed 20 times the dose of the polymer at the substrate.
These studies have focused on organic systems. However, some authors have sought to pattern inorganic materials using methods based on near-field microscopy. Madsen et al. used an NSOM coupled to an argon ion laser to write oxide structures into hydrogen-passivated silicon surfaces, which were then used as resists during etching of the Si substrate with KOH.[32] The structures observed were complex, but consisted of a central line with a full width at half maximum height of, in one case, 111 nm, and in another, 126 nm. These narrow structures were bordered by slightly wider features attributed to an interference pattern dominated by far-field excitation through the sidewalls of the uncoated fiber used. Narrower structures (50 nm) were observed in the absence of the optical excitation, attributed to the presence of the electrostatic potential between the probe and the amorphous Si layer. Significant improvements were produced when aluminum-coated fibers were used. The Al coating prevented far-field emission through the fiber walls, and resulted in the formation of much better defined structures with widths not much in excess of 50 nm.[33] These structures represent some of those most clearly resolved in these early studies.
Herndon et al. also used NSOM-based methods to fabricate structures in hydrogen-passivated silicon.[34] They explored wavelengths down to 248 nm, using argon ion and excimer lasers. The threshold dose required to fully expose the resist was found to decline with the wavelength of the light used. However, the lines written had widths that were comparatively large—in the range of 140-320 nm. Like Madsen et al., they also found that modification could occur in the absence of optical excitation.
Hosaka and coworkers examined a different inorganic system, amorphous Ge-Sb-Te films, whose structure could be modified by localized heating by light from a pulsed diode laser, emitting at 785 nm and delivered through an NSOM probe.[35,36] They achieved somewhat superior resolution, demonstrating the fabrication of 60 nm structures. A variety of other approaches have been explored, some of which are more complex and use less conventional resist materials. For example, Yamamoto et al. studied the deposition of zinc using the photo-dissociation of gas-phase dimethyl zinc.[37] While they reported the formation of sub-100 nm structures, these were comparatively ill defined. Better images were reported of 200 nm zinc structures. Hong et al. combined near-field microscopy with a micropipette filled with photoresist.[38] While their paper demonstrates an innovative modification of the method, they were only able to fabricate structures that were some hundreds of nanometers in size. Finally, Philipona et al. examined the direct attachment of biological molecules to photosensitive monolayers of a diazirine molecule.[39] This represented a significant step into a new area of application, but it proved only possible to create comparatively large (hundreds of nanometers) structures.
PHOTOCHEMISTRY OF SELF-ASSEMBLED MONOLAYERS
Summarizing the progress described above, it may reasonably be said that NSOM-based approaches had yielded new and, in some cases, unique capabilities, without yielding the kind of routinely very high resolution that might have been hoped in the earliest studies. The thickness of the resist layer has been demonstrated to be critical, with the potential for the spread of the excitation, through the divergence of the electric field, the diffusion of reactive species formed by photo-excitation, and thermal migration, increasing with film thickness. Many of these phenomena may be controlled through the selection of an appropriate resist material. In this section, we describe how the use of self-assembled monolayers (SAMs) of alkanethiols adsorbed on gold surfaces has provided unambiguous evidence not only that structures significantly smaller than 100 nm may be fabricated routinely, but also that near-field techniques can rival the power of electron beam lithography for such materials.
Alkanethiols, HS(CH2)nX, adsorb spontaneously onto gold surfaces, forming dense, well-ordered mono-layers, SAMs, that provide versatile templates for the construction of complex molecular architectures.[40] Li et al.[41] and Tarlov and Newman[42] reported that on extended exposure to the air, the alkylthiolate adsor-bate species was oxidized to yield an alkylsulfonate:
Huang and Hemminger[43] and Tarlov, Burgess, and Gillen[44] demonstrated that the same process could be initiated by exposure of SAMs to light from a mercury arc lamp. They found that unlike the adsorbates in the pristine monolayer, the sulfonate oxidation products were only weakly bound, and could be displaced, either by rinsing or by immersion in a solution of a second thiol, to generate a chemical pattern. The process is illustrated in Fig. 2, where a carboxylic acid-terminated thiol is photopatterned. After immersion of the sample in a solution of a methyl-terminated thiol, a chemical pattern is formed that may be imaged using a variety of methods including friction force microscopy (FFM). (Fig. 3), scanning electron microscopy (SEM)[45] and imaging secondary ion mass spectrometry (SIMS).[46] The advantage of depositing both chemistries in a solution-phase self-assembly process is that the entire specimen exhibits well-ordered adsorbates with relatively low defect densities.
The mechanism of photooxidation has been the subject of some debate. Early on, it was proposed by Hemminger and coworkers that the process was initiated by the formation of hot electrons at the gold surface. However, this idea did not have direct support. Studies in the authors’ laboratory demonstrated that for SAMs adsorbed on both gold[47] and silver surfaces,[48] the rate of oxidation decreased with increasing adsorbate chain length, correlating with the decrease in alkyl chain mobility associated with the increased dispersion interaction between long alkyl chains, and suggesting that penetration of oxygen species to the air-sulfur interface was the rate-limiting step. For adsorbates capable of forming hydrogen bonds between their terminal groups, the rate of oxidation was found to be even slower,[49] suggesting that the resulting network of hydrogen bonding interactions impeded the diffusion of oxygen. A radical proposal was made by Bohn and coworkers[50,51] and by Norrod and Rowlen,[52] who suggested that the process was not in fact a photochemical one, but involved the ozonolysis of adsorbate molecules. Critically, all of the studies published till date had utilized mercury arc lamps, which typically exhibited broad emission spectra. They suggested that short wavelength light generated by a mercury arc lamp could initiate ozone formation, and that ozone species could oxidize the adsorbate molecules in the SAM. In studies of the oxidation of hexadecanethiol SAMs, they claimed that the interposition of a filter that blocked short wavelength radiation between the lamp and the source extinguished the photooxidation process.[50]
Fig. 2 Schematic diagram illustrating the photopatterning of a SAM. A carboxylic acid-terminated SAM is formed and then exposed to UV light through a mask. In exposed areas, the adsorbates are oxidized to weakly bound alkylsul-fonates, which are displaced by a contrasting solution-phase thiol in the final step.
Fig. 3 An 80 mm x 80 mm FFM image of a patterned SAM. The pattern consists of regions functionalized by HS(CH2)uCH3(dark contrast) and HS(CH2)10COOH (bright).
This finding, if correct, would have meant that it will be difficult to utilize such approaches for near-field excitation, because a methodology based upon the creation of gaseous reagents would be unlikely to yield significant spatial control of reactivity. However, using a lamp equipped with a filter designed to eliminate ozone formation, Brewer et al. provided clear evidence that oxidation of SAMs could occur in the absence of ozone.[53] It was found that monolayers of carboxylic acid-terminated SAMs on both gold and silver surfaces oxidized rapidly to yield alkylsulfonates when exposed to light with a wavelength of 254 nm. Recently, a more detailed kinetic study has been carried out using static SIMS, to determine rates of photooxidation.[54] Significant differences were observed between the rates of oxidation under these well-defined conditions and those measured using mercury arc lamp sources. Unexpectedly, when exposed to 254 nm light, acid terminated thiols oxidize much faster than methyl-terminated adsorbates—the direct opposite of the observation made earlier for an arc lamp source. This is explained by differences in the mechanisms of oxidation. The dominance of intermolecular interactions in determining the rate of oxidation during exposure to the arc lamp source suggests that gaseous reagents are important under those conditions, in agreement with the predictions of Bohn and coworkers.[50] However, when a well-defined light source emitting at 254 nm is used, these species are absent, and it is believed that hot electron formation provides the main impetus for SAM oxidation. Measurements of the contact potential differences of SAMs on Au and Ag have yielded data that reveal differences in SAM work functions that are consistent with such an explanation.[54] In particular, it was found that the work functions of carboxylic acid-terminated SAMs are larger than the energy of a 254 nm photon. As a result, absorption of a UV photon may lead to the promotion of an electron from the Fermi level to an excited state, but not to photoemission. These excited electrons may tunnel into an antibonding state in the adsorbate sulfur atom, leading to oxidation. In contrast, the work functions of methyl terminated SAMs are smaller than the photon energy, meaning that photoemission may occur. Photoemitted electrons will have small kinetic energies, and most likely be scattered by the alkyl chains above the surface, but nevertheless the number of electrons available for the initiation of oxidation will be significantly reduced. The transfer of hot electrons from the Fermi level to the adsorbate sulfur thus becomes the rate-limiting step, although adsorbate order still plays a role. For example, carboxylic acid-terminated thiols pack less closely on silver than on gold, and a concomitant increase was reported in the rate constants for photooxidation.
