© 1999 Robert A. Freitas Jr. All Rights Reserved.

Robert A. Freitas Jr., Nanomedicine, Volume I: Basic Capabilities, Landes Bioscience, Georgetown, TX, 1999

2.4.1 Molecular Mechanical Components

In order to lay a foundation for molecular manufacturing, it is necessary to create and to analyze possible designs for nanoscale mechanical components that could, in principle, be manufactured. Because these components could not yet be built in 1998, such designs could not be subjected to rigorous experimental testing and validation. Designers were forced instead to rely upon ab initio structural analysis and computer studies including molecular dynamics simulations. Noted Drexler:¹⁰ "Our ability to model molecular machines (systems and devices) of specific kinds, designed in part for ease of modeling, has far outrun our ability to make them. Design calculations and computational experiments enable the theoretical studies of these devices, independent of the technologies needed to implement them."

In nanoscale design, building materials do not change continuously as they are cut and shaped, but rather must be treated as being formed from discrete atoms.²⁸⁴⁵ A nanoscale component is a supermolecule, not a finely divided solid. Any stray atoms or molecules within such a structure may act as dirt that can clog and disable the device, and the scaling of vibrations, electrical forces, thermal expansion, magnetic interaction and surface tension with size lead to dramatically different phenomena as system size shrinks from the macroscale to the nanoscale.¹⁰

Molecular bearings are perhaps the most convenient class of components to design because their structure and operation is fairly straightforward. One of the simplest examples is Drexler's overlap-repulsion bearing design,¹⁰ shown with end views and exploded views in Figure 2.28 using both ball-and-stick and space-filling representations. This bearing has exactly 206 atoms including carbon, silicon, oxygen and hydrogen, and is comprised of a small shaft that rotates within a ring sleeve measuring 2.2 nm in diameter. The atoms of the shaft are arranged in a 6-fold symmetry, while the ring has 14-fold symmetry, a combination that provides low energy barriers to shaft rotation.* Figure 2.29 shows an exploded view of a 2808-atom strained-shell** sleeve bearing designed by Drexler and Merkle¹⁰ using molecular mechanics force fields to ensure that bond lengths, bond angles, van der Waals distances, and strain energies are reasonable. This 4.8-nm diameter bearing features an interlocking-groove interface which derives from a modified diamond {100} surface. Ridges on the shaft interlock with ridges on the sleeve, making a very stiff structure. Attempts to bob the shaft up or down, or rock it from side to side, or displace it in any direction (except axial rotation, wherein displacement is-extremely smooth) encounter a very strong resistance.²⁷⁹ Whether these bearings would have to be assembled in unitary fashion, or instead could be assembled by inserting one part into the other without damaging either part, had not been extensively studied or modeled by 1998.

* At the atomic scale, the two opposing surfaces have periodic bumps and hollows, but the periods of these bumps are different for the two surfaces -- that is, they are "incommensurate".^10,3243 Two incommensurate surfaces cannot lock up in any particular position, hence the barrier to free rotation is very low, on the order of ~0.001 kT (thermal noise at room temperature).²⁸⁰

** Components of high rotational symmetry may consist of (a) intrinsically curved, (b) strained-shell, or (c) special-case structures.¹⁰ In the case of (b), the bearing illustrated in Figure 2.29 has bond strains of around ~10% (~38 zJ/atom), and similar hydrocarbon bearings have been designed with bond strains of ~5% (~11 zJ/atom) [R. Merkle, personal communication, 1998]. For comparison, strain energies^1866,2615 are <~3 zJ/atom for diamond lattice, ~25 zJ/atom in C₂₄₀, ~7-27 zJ/atom in the walls of infinite carbon nanotubes of diameter 0.7-1.3 nm, up to ~59 zJ (13% strain) for some bonds around a Lomer dislocation in diamond,²⁸⁶¹ ~70 zJ/atom in C₆₀, and at least ~80 zJ/atom for C₃₆. Fullerenes are among the most highly strained natural molecules ever isolated. For symmetrical diamondoid structures with negligible hoop stress, permissible bond strains may in theory be as large as ~140 zJ/atom producing a ~23% bond strain;¹⁰ nanotube breaking strain is 20-30% for various chiral forms, and buckling strain is ~8% in axial compression. Bond strain in a simple strained-shell bearing can be lowered by making the bearing bigger, thereby reducing the curvature. Thus strained shell bearings are feasible, although in 1998 it remained unclear exactly how small they could be before becoming unstable.

Molecular gears are another convenient component system for molecular manufacturing design-ahead. For example, Drexler and Merkle¹⁰ designed a 3557-atom planetary gear, shown in side, end, and exploded views in Figure 2.30. The entire assembly has twelve moving parts and is 4.3 nm in diameter and 4.4 nm in length, with a molecular weight of 51,009.844 daltons and a molecular volume of 33.458 nm³. An animation of the computer simulation shows the central shaft rotating rapidly and the peripheral output shaft rotating slowly. The small planetary gears rotate around the central shaft, and they are surrounded by a ring gear that holds the planets in place and ensures that all of the components move in the proper fashion. The ring gear is a strained silicon shell with sulfur atom termination; the sun gear is a structure related to an oxygen-terminated diamond {100} surface; the planet gears resemble multiple hexasterane structures with oxygen rather than CH₂ bridges between the parallel rings; and the planet carrier is adapted from a Lomer dislocation²⁸⁹⁴ array created by R. Merkle and L. Balasubramaniam, and linked to the planet gears using C-C bonded bearings. View (c) retains the elastic deformations that are hidden in (a) -- the gears are bowed. In the macroscale world, planetary gears are used in automobiles and other machines where it is necessary to transform the speeds of rotating shafts.

W. Goddard and colleagues at CalTech^2844,2845 performed a rotational impulse dynamics study of this "first-generation" planetary gear. At the normal operational rotation rates for which this component was designed (e.g., <1 GHz for <10 m/sec interfacial velocities), the gear worked as intended and did not overheat.²⁸⁴⁴ Started from room temperature, the gear took a few cycles to engage, then rotated thermally stably at ~400 K. However, when the gear was driven to ~100 GHz, significant instabilities appeared although the device still did not self-destruct.²⁸⁴⁴ One run at ~80 GHz showed excess kinetic energy causing gear temperature to oscillate up to 450 K above baseline.²⁸⁴⁵ One animation of the simulation shows that the ring gear wiggles violently because it is rather thin. In an actual nanorobot incorporating numerous mechanical components, the ring gear would be part of a larger wall that would hold it solidly in place and would eliminate these convulsive motions which, in any case, are seen in the simulation only at unrealistically high operating frequencies.

Drexler and Merkle²⁸⁴⁷ later proposed a "second-generation" planetary gear design (Fig. 2.31) with 4235 atoms, a molecular weight of 72,491.947 daltons and a molecular volume of 47.586 nm³. This new version was indeed more stable but still had too much slip at the highest frequencies. Commenting on the ongoing design effort, Goddard²⁸⁴⁵ suggested that an optimal configuration could have the functionality of a planetary gear but might have an appearance completely different from the macroscopic system, and offered an example: "Because a gear tooth in the xy plane cannot be atomically smooth in the z-direction, we may develop a Vee design so that the Vee shape of the gear tooth in the z-direction nestles within a Vee notch in the race to retain stability in the z-direction as the teeth contact in the xy plane. This design would make no sense for a macroscopic gear system since the gear could never be placed inside the race. However, for a molecular system one could imagine that the gear is constructed and that the race is constructed all except for a last joining unit. The parts could be assembled and then the final connections on the face made to complete the design" analogous to the ZARBI system of Rebek.¹³¹

Another class of nanodevice that has been designed is a gas-powered molecular motor or pump.²⁸⁵⁸ The pump and chamber wall segment shown in Figure 2.32 contain 6165 atoms with a molecular weight of 88,190.813 daltons and a molecular volume of 63.984 nm³. The device can serve either as a pump for neon gas atoms or (if run backwards) as a motor that can convert neon gas pressure into rotary power. The helical rotor has a grooved cylindrical bearing surface at each end, supporting a screw-threaded cylindrical segment in the middle. In operation, rotation of the shaft moves a helical groove past longitudinal grooves inside the pump housing. There is room enough for small gas molecules only where facing grooves cross, and these crossing points move from one side to the other as the shaft turns, moving the neon atoms along. Goddard²⁸⁴⁵ reported that preliminary molecular dynamics simulations of the device showed that it could indeed function as a pump, although "structural deformations of the rotor can cause instabilities at low and high rotational frequencies. The forced translations show that at very low perpendicular forces due to pump action, the total energy rises significantly and again the structure deforms." Merkle acknowledged that the pump moves neon atoms at an energy cost of 185 Kcal/mole-Angstrom (12,900 zJ/atom-nm), which is not very energy-efficient. Further refinement of this crude design is clearly warranted.

Conveyor systems would also be useful. Employing a primitive molecular CAD software package called Crystal Sketchpad, G. Leach²⁸⁶¹ created a design for a 5-nanopart, ~2500-atom conveyor belt system comprised of two rollers, two axles, and a belt consisting of a strained thin-walled diamond sheet. The design has not been subjected to further computational analysis, either to minimize rotational energy barriers or to optimize rotational dynamics or operational stability.

Drexler and Merkle²⁸⁵⁹ have also produced a preliminary design for a 2596-atom fine-motion controller (Fig. 2.33). A general-purpose molecular assembler arm must be able to move its "hand" by many atomic diameters, position it with fractional-atomic-diameter accuracy, and then execute finely-controlled motions, perhaps to transfer one or a few atoms in a guided chemical reaction. Human arms use large muscles and joints for large motions and more finely-controlled finger motions for precision; the device presented here can execute precise finger-like motions over several atomic diameters with associated 90-degree rotations. The core of the device consists of a shaft linking two hexagonal endplates, sandwiching a stack of eight rings, making a modified Stewart platform (Section 9.3.1.5). In a complete system, each ring would be rotated by a lever driven by a cam mechanism. Each ring supports a strut linked to a central platform (here shown raised, displaced, and twisted). Rotating a ring moves a strut; moving a strut moves the platform; positioning all eight rings (over)determines a platform position in x, y, z, roll, pitch, and yaw. (If the struts were rigid, six would do the job; here, two struts have been added to increase stiffness.) Notes Drexler: "The chief design problem is to enable an adequate range of motion without mechanical interference or unacceptable bond strains, and within the size constraints set by available modeling tools and patience."

Almost all current design research in molecular nanotechnology is restricted to computer simulation, which allows the design and testing of large structures or complete nanomachines, and the compilation of growing libraries of molecular designs. The work is relatively inexpensive and does not require the support of a large team. Of course, calculations of many-body systems are notoriously difficult, with many computer packages making a number of simplifying assumptions -- e.g., nuclei as point masses, electrons treated as a continuous charge distribution, and 3-D potential energy functions derived semi-empirically from experimental data and treated as a classical field despite their true quantum mechanical character (for ease of computation). Notwithstanding these shortcuts, Tuzun and colleagues²⁶⁶⁴ claim that classical simulations may overestimate the rate of energy transfer between vibrational modes even at low energies, in which case "current designs for various nanocomponents [would] actually perform better and be more stable than recent molecular dynamics simulations suggest."

Goddard²⁸⁵³ notes that future nanosystem simulations may require 1-100 million atoms to be considered explicitly, demanding major improvements in molecular dynamics methodologies. By 1998, new algorithms for parallel processing of massive molecular dynamics simulations were being developed,²⁸⁵⁴ producing methods and optimized parallelized computer programs efficient for high capacity molecular dynamics simulations of 10,000-1,000,000 atoms for finite molecular structures.

Ultimately, Computer-Aided (Molecular) Design (CAD) systems²⁸⁶¹ will be needed to efficiently design and analyze molecular components and their higher-order assemblies. CAD systems are commonplace in macroscale engineering and architecture, as are Computer-Aided Manufacturing (CAM) systems in macroscale manufacturing. In the molecular realm, Computer Aided Synthesis Design (CASD)* has been avidly pursued by computational chemists since the 1980s, and by 1998 many university and commercial molecular modeling software packages were already well-known, including Alchemy (Tripos Inc.; www.tripos.com), Cerius2 (Molecular Simulations Inc.; www.msi.com), Chem3D (CambridgeSoft; www.camsoft.com), Conformer (Princeton Simulations; www.conformer.com), Gaussian94 (Gaussian Inc.; www.gaussian.com), Hyperchem (Hypercube Inc.; www.hyper.com), Molecular Operating Environment (Chemical Computing Group Inc.; www.chemcomp.com), MOPAC97 (Fujitsu Ltd.; www.winmopac.com), RealMol/CAVE/NAMD (Fraunhofer Institute for Computer Graphics; www.igd.fhg.de/www/igda4/research/chemie), Sculpt (Interactive Simulations Inc.; www.intsim.com/products/index.htm), and Spartan (Wavefunction Inc.; www.wavefun.com).

* Existing CASD software/database packages include CAMEO (reaction chemistry assistant; William L. Jorgensen, Yale Univ.; zarbi.chem.yale.edu/programs/cameo.html), CAS (reaction data base; www.cas.org/CASFILES/casreact.html), Chiron (retrosynthetic analysis of chiral precursors; Steve Hanessian, Univ. of Montreal; www.netsci.org/Resources/Software/Cheminfo/chiron.html), CIARA (Vogel Scientific Software Inc.; www.vogelscientific.com), Crossfire (Computer Assisted Synthesis, Beilstein Information Systems; www.beilstein.com), EROS (Johann Gasteiger, Erlangen University; schiele.organik.uni-erlangen.de), LAHSA (retrosynthetic analysis; Alan Long, Harvard University; long@midas.harvard.edu), REACCS (MDL Information Systems Inc.; www.mdli.com), SYNGEN (Jim Hendrickson, Brandeis University; syngen2.chem.brandeis.edu/syngen.html), SYNLIB (W. Clark Still, Columbia University), and SYNTREE (Trinity Software Inc.; www.trinitysoftware.com/Trinity/orgchem/SYNTREE.HTM).

In 1998, the capacity of such systems to design and manipulate large nanoscale mechanical components was extremely limited. A few very primitive molecular nanotechnology design packages had been attempted, including Crystal Sketchpad,²⁸⁶¹ DiamondCAD (www.zyvex.com/diamond.html), Molecular Assembly Sequence Software (www.carol.com/mass.shtml), Molecular Modelling Toolkit (starship.python.net/crew/hinsen/mmtk.html), and NanoCAD (world.std.com/~wware/ncad.html). The possible design of molecules for specific purposes using genetic software techniques had also been investigated.²⁹⁵⁵

Last updated on 7 February 2003