John
M. Michelsen, Mark J. Dyer, and Jim Von Ehr
Zyvex LLC 251 West Renner Parkway, Suite 166
Richardson, TX 75080
This was presented at the Fifth Foresight Conference on Molecular Nanotechnology.
NOTE: This page uses the <sup> and <sub> convention for superscripts and subscripts. If "103" appears as "103" then your browser does not support superscripts. If "xi" = "xi" then subscripts are not supported.
Zyvex LLC is a developmental engineering company whose goal is to develop an assembler, a nanometer scale machine capable of manufacturing a wide variety of atomically precise structures via mechanochemistry. Although theoretical investigations indicate that such machines are possible and can be designed in atomic detail, the capabilities and operations necessary to construct them are less clear. Present-day scanning probe microscopes, while limited to imaging and rudimentary pick-and-place assembly operations, appear to be necessary tools for developing the next generation of more sophisticated assemblers. This paper addresses the construction of covalently bonded nanometer-scale components using existing scanning tunneling microscope(STM) technology. Schemes for three-dimensional, atomically precise construction in ultrahigh vacuum (UHV) are presented through the use of scaffolding, intermediate components, and interface joiners.
Exploratory engineering of molecular systems in the past decade has produced a number of atomic scale designs for mechanical devices, proposed as components of atomic scale machine systems (Drexler, 87 and 92). More recently, ab initio chemical modeling of atomic addition steps (Walch and Merkle, 97) has been pursued with the goal of verifying or rejecting deposition tools and sequences proposed in the construction of such devices. These studies assume a very high degree of spatial control over the deposition tool, which with the evolution of a wide class of scanning probe microscopes (SPM) is becoming experimentally feasible. SPM's provide several of the essential features of an atomic scale construction device including sub-angstrom positioning accuracy, compatibility with UHV environments, and near real-time imaging of atomic modifications. The latter quality continues to improve, and frame update speeds are now approaching standard video rates for tunneling microscopes (Curtis et al. 97). Two largely unexamined topics in molecular manufacturing research today, however, are the adaptation of real-world proximal probe manipulation operations to the atomically precise construction motions required for building molecular machines, and construction sequences (SPM-based or otherwise) for such basic components as the sliding interfaces prevalent in most proposed assembler component designs (Merkle, 96). Sequences for the manufacture of simple building blocks, and eventually bearings, actuators and other nanoscale mechanical articulating components via proximal probe are necessary steps toward constructing an entire assembler with SPM technology.
A prime constraint in SPM-based construction of non-bonded interfaces is the dimensional limitation of a single positioning probe's operation. Unlike conventional robotic manufacturing where multiple arms can fit objects into interfaces in three dimensions, objects made by present-day SPM's must be mostly layered constructions. Similar construction obstacles have been addressed in autofab laser resin curing, where scaffolding is built up around movable parts to immobilize them (Burns, 93). To produce interfaces where the space between parts is measured in tenths of an angstrom, an analogous approach would be to insert and remove sacrificial supports and interface joiners, or molecular "spacers," during the construction process. This process requires the reproducible picking and placing of atoms via SPM to construct the spacers along interfaces, in addition to building and placing the interfaced components themselves if they have not been fabricated elsewhere and guided into position by other means. While mature molecular manufacturing systems may resemble conventional robotic factories in many respects, SPM placement of non-bonded parts into closely-spaced interfaces requires bonding them in place using joiners; this precludes the use of angular probe motions or assistance by additional moving parts such as robotic grippers.
Among the techniques necessary for assembler construction is a basic atomic addition procedure. In the simplest case, assembly of covalent structures proceeds by the addition of only one atom type, though numerous covalent two component materials are possible. The set of candidate one-component covalent materials includes the elements carbon, silicon, and perhaps germanium; the construction substrates under consideration can also be surfaces of diamond, silicon, or germanium, respectively, to ensure that such constructions can be extended into three dimensions. What remains to be specified in a mechanosynthesis approach is the process by which individual atoms are to be recognized and transferred, the number of transfer steps necessary per atom, and any materials on which the atoms are to be adsorbed intermediate to their initial state and final site in a crystal lattice.
Of the aforementioned covalently-bonded surfaces, pick and place operations have been demonstrated on such surfaces as Ge(111)-2x8, Si(111)-7x7 and Si(100)-2x1 by Becker et al. 87, Aono et al. 93 and Avouris 95, among others. Under various applied fields and tip-sample distances, small groups or individual atoms were found to be transferred from the substrate to a tungsten STM tip in vacuum, and could thereafter be redeposited on the substrate by field evaporation.
Figure
1.
(a) A silicon atom triply bonded to the apex of a tungsten STM tip, over a portion of the Si(100)-2x1 reconstructed surface.
(b)
A silicon atom bonded to three atoms at the apex of a STM tip.
Diffusion away from the STM apex during or after loading of the tip compromises
site-specific lattice addition.
In order to add reliably a single atom to a specific lattice point, the atom must be relatively immobile at the apex of a STM tip during sample approach. A sequence of events proposed to explain the results of silicon atom deposition from a W(111) STM tip (Grey et al. 94) includes the field-assisted migration of previously adsorbed atoms to the tip apex prior to evaporation. When the migrating atom experiences a high enough electric field, it leaves the tip. Since migration, evaporation and deposition are usually driven by a single voltage pulse, deposition site uncertainties can arise not only from uncertainties in the position and momentum of the atoms immediately before an atom leaves the tip, but also from uncertainty in the atom's location with respect to the tip apex prior to the applied voltage pulse. Ideally, diffusional motion should cease and the stationary atom should reside at the tip apex before evaporative or chemically-driven deposition occurs. Figure 1 illustrates two locations a silicon atom can occupy as the tunneling atom of a STM tip. The imprecision of single atom placement experiments to-date can be explained in part by uncertainties related to atom migration and tip structure, but also by noting that field evaporation in picking adatoms from the Si(111)-7x7 surface breaks three silicon-silicon bonds (~169.5 kcal/mol) and releases further vibrational energy upon the formation of tungsten-silicon bonds. The diffusional barrier for silicon adatoms on tungsten is very low (Casanova and Tsong, 82) at 0.7 eV or 15.5 kcal/mol on W(110). There is therefore little chance the desorbed atom will stay on the apex of the tip even in chemically-assisted field evaporation, where only a third of the threshold electric field strength reported for field ionization microscopy studies is required for desorption via STM (Lyo and Avouris, 91).
The transfer of silicon atoms from a specific site to a STM tip and then to the Si(100)-2x1 surface (which has a simpler reconstruction and better atom stacking properties than Si(111)-7x7, with the same diffusional barrier), should require transfer steps each leaving the adatom at its new adsorption site with less than ~10 kcal/mol vibrational energy. Chemical driving forces in the transfer of a Si atom from gold to tungsten to silicon appear to meet these requirements. While the differences in adsorption energies for the gold/tungsten and tungsten/silicon transfer steps are uncertain and may not fall within 10 kcal/mol, when the Si atom is halfway between surfaces it is bonded to both and the system is in a local energy minimum. Any vibrational energy released by bonds formed will thermalize to the tip and sample on a time scale much shorter than that of the tip-sample contact. In field evaporation, however, when the Si atom is halfway between surfaces it is not fully bonded to either, and the system is on an energy maximum. When the atom impacts its new site it bears both kinetic energy and energy released upon forming new bonds, and will diffuse or hop from the tip apex before the energy is thermalized (Wolkow, 95).
Experimentally testing this gold/tungsten/silicon transfer sequence should be straightforward. Thin film gold "islands" could be sputtered through a lacey carbon transmission electron microscope grid onto a clean Si(100)-2x1 surface followed by the deposition of roughly 1% of a monolayer of silicon atoms. This process would yield relatively small gold reservoirs of adsorbed silicon adatoms adjacent to clean Si(100)-2x1 spots for a STM to access. At low temperatures, a tungsten STM tip could then pluck individual silicon adatoms from the gold islands, relocate to the nearby clean Si(100)-2x1 surface, and deposit them at specific lattice points using only chemical forces (Meyer and Rieder, 97).
Another transfer scheme is the placing of carbon atoms from surface-bound deposition tools. In this method the scanning probe tip is used as the imaged substrate, i.e. it may not be atomically sharp, and an array of projections on the substrate serve dual purposes as atom-bearing deposition tools and imaging probes (Drexler, 91). For example, 1% of a monolayer of an activated carbene deposition tool (Walch and Merkle, 97) could be cyclo-added to highly oriented pyrolytic graphite from the gas phase (Figure 2a). Transfer of carbon atoms to a carbon-coated tip could then be accomplished in the same manner as with picking and placing, in that the tip approach would be performed when the desired transfer location above the tool was imaged. If the rotational orientation of the tool with respect to the substrate is such that the bonds formed disrupt the prior tool pi-bond, carbon deposition should be possible without the twisting motion previously proposed. Furthermore, similar deposition strategies should be possible with other atoms that form strong pi-bonds with carbon, such as oxygen and nitrogen, and on most surfaces that form strong sigma bonds such as silicon, since two bonds are formed and only one broken on tip-sample separation.
 |
 |
 |
 |
Figure 2.
(a) A carbon deposition tool bonded to a graphite substrate, facing a carbon-coated
STM tip.
(b), (c), and (d) Transformation of orbitals in the deposition sequence.
For carbon adatoms on diamond(100), the most favorable adsorption site (forming a three-membered ring across a dimer) is 56 kcal/mol lower in energy than the next most favorable site bridging two dimers by B3LYP/6-31G(d) density functional single-point energy calculations with Gaussian 94. Rearrangement and migration of carbon adatoms from the dimer-adsorbed site on diamond(100) therefore seems unlikely.
Construction of any three dimensional shape will depend critically on the stacking properties of the atoms used with respect to the atomic layer below. For instance, on the Si(100)-2x1 surface the local energy minimum of an adsorbed Si adatom, highlighted in Figure 3a, is to one side and between dimers along a dimer row (Brocks et al. 93). Deposition of another atom next to the first forms a new dimer as shown in Figure 3b, which has the alternative rotated (epitaxal) configuration (Figure 3c), 0.08 eV or 1.9 kcal/mol lower in energy (Yamasaki and Uda, 96). Exploiting this more stable arrangement, an array of isolated dimers can be deposited in parallel, illustrated in part in Figure 3d. Further deposition could proceed between dimers to form the U-shape shown in Figure 3e, then in the remaining unoccupied site between dimers to complete a new dimer row aligned perpendicular to those in the layer below (Figure 3f). On carbon(100), which forms a reconstruction similar to the analogous silicon surface but with stronger pi-bonds, the initial adatom would be located directly on top of a dimer, and deposition of the next atom would therefore proceed directly to the epitaxal dimer stage (Stallcup et al. 95)
 |
 |
 |
 |
 |
 |
Figure 3.
(a)
An absorbed Si atom on the Si(100)-2x1 surface.
(b) Initial dimer conformation after new Si atom addition to the right of
the previously deposited atom.
(c) The more stable epitaxally oriented dimer conformation.
(d) Two neighboring dimers,
(e) single-atom bridged dimers, and
(f) a new dimer row. On diamond(100) non-epitaxal dimer creation is unnecessary,
but the last three steps are identical.
Scaffolding and differential etching reactions are valuable techniques for creating complex features and movable components on macro- and microscopic scales, so it seems reasonable to explore the use of such methods on an atomic scale to form mechanical devices. As in conventional microfabrication or stereolithography, proximal probe constructions must do without the complex multiple-arm grasping, rotating, and placing motions usually associated with constructing and fitting components together to make complex machines. For atomically precise fabrication and tolerances, there is little room between tightly fitted parts for a sacrificial scaffolding material, except in limited circumstances where the scaffolding bridges an non-critical interface, or reciprocal surfaces become terminated upon removal of the scaffold and the interface closes. For example, if a few atomic layers of germanium are interposed between oxygen-terminated structural silicon parts, such as in the slide-and-race construction depicted in Figure 4, an aqueous hydrochloric acid etch could in principle remove the germanium and further terminate the silicon lattice, provided that sufficient space exists for the diffusion of released germanium species. In an aqueous environment, however, the etch must proceed in a controlled manner such that the scaffolding is removed without permitting the slide to break away from the race, where it could be directed and captured by van der Waals attractive forces. In this scenario, the oxygen termination is chosen to shield the silicon parts from the etchant. Vacuum construction and etching of silicon-scaffolded diamond components with chlorine may be an attractive alternative; however, silicon has a lattice constant much larger than diamond, and so it would have to be deposited in a non-equilibrium form. For carbon-based constructions, chlorine is also chosen as an etchant because it is unlikely to disrupt the diamond lattice, but it will terminate the diamond at the vacated silicon attachment points.
 |
 |
Figure 4.
(a)
Oxygen-terminated silicon component (purple and green) with germanium scaffolding
(dark purple).
(b) Resulting structure after etching of germanium.
The use of etching reactions and the resulting interruptions of the construction process could be avoided if intermediate components and joiners could be fabricated elsewhere, by SPM or other means, then harvested and placed or removed with a proximal probe. As in the scaffolding example, the key issue in placing components to function as movable parts is the fabrication of non-bonded interfaces. Following the pick-and-place strategy for construction, components could also be transferred between the same three materials used for atomic transfer. For example, if the initial atom stacking proceeds on a small island of silicon weakly adsorbed on gold, the resulting component could be picked off the gold surface in its entirety by a tungsten STM tip, or by a tip coated with strategically positioned intermediate joiners to maintain component orientation. If the now exposed surface of the silicon component is satisfied with lattice termination moieties, a part could be placed from the STM tip into a prepared interface site with yet stronger bonds to silicon. A slider such as that pictured in Figure 4 can thus be assembled without scaffolding, provided the adsorption energies were weak relative to the internal bonding of the component. For instances where the tip-part adhesion exceeds that of van der Waals forces between interfaced systems, such as slider-channel geometries, combined sheer and normal motions of the probe can sever the tip from the part, provided sufficient clearance around the components exists.
 |
 |
 |
 |
 |
Figure 5.
(a)
Cutaway views of a Si2 (purple) joiner-held block of hydrogen-terminated
diamond(100),
shown in white and red.
(b) After addition of one atomic layer of carbon.
(c) The result of removing the Si2 joiners from the right-hand
side, addition of a layer of carbon,
and re-application of Si2 joiners.
(d) After the same procedure on the left-hand side.
(e) The resulting structure after addition of another atomic layer of carbon.
Figure 5 illustrates an alternative building process in which a slider and channel are built up togther in layers, with interface joiners serving to immobilize and align the slider as layers are added. Interface joiners, such as those used to transfer a component to the substrate in a placing operation, should not be made up of a substantial amount of material so that they can be readily and cleanly removed, and they should be nearly level with the working surfaces to avoid conflicting with nearby probe operations. A linear diatomic molecule about the length of the nonbonded radii of two surface termination moieties is ideal, with the joiner not having a great affinity for the surface termination moiety to make that abstraction unlikely. Candidate joiners are the S2 and Si2 dimers, which could be formed by bringing together two radicals terminating the slide and the race in component placement, or by adding one atom at a time in the deposition tool array strategy. They would be removed by voltage pulses or abstraction tools and added again continuously as the structure is built higher, as shown in Figure 5 for a diamond component.
Displacements of tens of nanometers or more are not usually achieved with sub-angstrom precision; therefore, transporting components will be problematic if the drift and hysteresis suffered by most piezo-based SPMs are not corrected at sub-nanometer levels by some form of feedback control. Low temperatures, probe/stage miniaturization and more repeatable electrostrictive actuators may address this to some extent, but another means actively to overcome this difficulty is to provide continual error correction during component placement. This can be accomplished by tracking the component position through inverse imaging with substrate probes or features of known location, and developing algorithms rapidly to recognize interactions between the characterized component-bearing tip and targeted deposition site. As mentioned earlier, the substrate imaging probes can also serve as tool arrays and provide surface termination of an exposed component face.
Two further strategies have no existing counterpart but are unique to the properties of small molecular systems and the limitations of proximal probes. The first is to design and build components without any non-bonded sliding surfaces parallel to the substrate. All net horizontal interfaces then would take the form of a "V", with construction of the bottom of the V beginning with interface joiners holding a line of atoms off the substrate. Construction would proceed by removal of a small fraction of the joiners, deposition of material on both sides of an angled interface, and addition of new joiners as in Figure 5. The considerations of joiner atomic structure are similar to those in the component placing strategy, except that angled interfaces present a larger gap to fill than horizontal ones and so will require longer joiners. If strictly layer-by-layer construction of parts is pursued, the manufacture of sliding interfaces that are angled with respect to the substrate, in order to enclose parts or for other reasons, will require more substantial joiners than diatomic species.
A second strategy involves the creation of non-bonded interfaces via construction on tip step-edges perpendicular to the substrate, as shown in Figure 6. The deposition tool shown is taller than that previously proposed, to allow carbon atom placement past three vertical atomic layers of the held component. Alternatively, the deposition tool could be attached to the end of a functionalized nanotube for larger, more complex constructions. The carbon atom to be deposited also projects at a sixty degree angle to the substrate, so that it can act as the tunneling atom both in conventional STM imaging and as a side-wall imager while minimizing tip/sample convolution effects. The topmost carbon is not a carbene as previously depicted, but is a sp2 radical bearing a hydrogen, both of which are deposited. The placing motion and transformation of orbitals, however, is similar to that depicted in Figure 2, using two radicals on the surface to break the carbon-tool pi bond thereby allowing exothermic deposition of carbon and hydrogen. A nitrogen or oxygen atom also could become the non-bonded surface moiety, and could be deposited from a tool bearing a nitrogen radical or ketone, respectively. For the layers farther away from the non-bonded interface and closer to the deposition tool, an angled carbene carbon deposition tool would be used. The diamond surface on which construction proceeds is the (110) surface, at a ninety degree angle to the (100)-2x1. This is the surface originally proposed for use of the carbene tool in Drexler, 92.
Figure 6.
Use
of a tall CH deposition tool to create horizontal non-bonded interfaces on
diamond.
The component is held on two or more sides by removable joiners, not shown
in this cut-away view.
The construction of a complete assembler by STM will require the production of components having non-bonded interfaces, such as bearings, actuators, and other articulating components. Further details of the procedures described here, using scaffolding and joiners, imaging and placement schemes, and various substrate constructions, will depend on the types of atoms employed, how they are to be applied and in what sequence. More exploration of proximal-probe atom and component placement processes with adsorbed carbon, silicon, and atoms of lower valence, of molecules strongly and stiffly bonded to surfaces for deposition tools and of selective etching and STM voltage desorption reactions will undoubtedly lead to heretofore unexamined combinations and structures. For practical nanometer-scale machine systems, considerable experimental work in these areas is necessary to bridge the gap between existing theoretical and experimental achievements.
Aono, M., Kobayashi, .A, Grey, F., Uchida, H. and Huang, D.H. (1993) "Tip-sample interactions in the scanning tunneling microscope for atomic-scale structure fabrication," J. Appl. Phys. 32, 1470-1477 and references therein.
Avouris, P (1995) "Manipulation of matter at the atomic and molecular levels," Acc. Chem. Res. 28, 95-102 and references therein.
Becker, R.S., Golovchenko, J.A. and Swartzentruber, B.S. (1987) "Atomic-scale surface modifications using a tunneling microscope," Nature 325, 419-421.
Brocks, G., Kelly, P.J. and Car, R. (1991) "Binding and diffusion of a Si adatom on the Si(100) Surface," Phys. Rev. Lett. 66, 1729-1732.
Burns, M. (1993) Automated Fabrication, Prentice-Hall. * Casanova, R. and Tsong, T.T. (1982) Thin Solid Films 93, p. 41.
Curtis, R., Toshiyuki, M. and Ganz E. (1997) "An ultrahigh vacuum high speed scanning tunneling microscope," Rev. Sci. Instrum. 68, 2790-2796.
Drexler, K.E. (1987) "Nanomachinery: Atomically precise gears and bearings," IEEE Micro Robots and Teleoperators Workshop, Hyannis, Massachusetts. IEEE, Cat. # 87TH0204-8.
Drexler, K.E. (1991) "Molecular tip arrays for molecular imaging and nanofabrication," J. Vac. Sci. Technol. B 9, 1394-1397.
Drexler, K.E. (1992) Nanosystems: Molecular machinery, manufacturing, and computation, Wiley.
Grey F., Huang, D. and Aono, M. (1994) "The sound of one atom hopping: Atomic manipulation on silicon surfaces by STM," Phil. Mag. B. 70, 711-720.
Lyo, I.W. and Avouris, P. (1991) "Field-induced nanometer- to atomic-scale manipulation of silicon surfaces with the STM," Science 253, 173-176.
Merkle, R.C. (1996) "Design considerations for an assembler," Nanotechnology 7, 210-215.
Meyer, G. and Rieder, K.H. (1997) "Controlled manipulation of single atoms and small molecules with the scanning tunneling microscope," Surface Science 377-379, 1087-1093.
Stallcup, R.E., Aviles, A.F., and Perez, J.M. (1995) "Atomic resolution ultrahigh vacuum scanning tunneling microscopy of epitaxal diamond (100) films," Appl. Phys. Lett. 66, 2331-2333.
Walch, S.P. and Merkle, R.C. (1997) "Theoretical studies of diamond mechanosynthesis reactions"
Wolkow, R.A. (1995) "Evidence of adsorbed atom pairing during homoepitaxial growth of silicon," Phys. Rev. Lett. 74, 4448-4451.
Yamasaki, T. and Uda, T. (1996) "Initial process of Si homoepitaxial growth on Si(001)," Phys. Rev. Lett. 76, 2949-2952.
Figures produced with Hyperchem 5.0
