IMM Report Number 10

Mechanosynthesis Simulations

Mechanosynthesis is a key part of the capabilities expected from mature nanotechnology. The positioning of reactants so that they bond to one site, but not to another chemically equivalent site a nanometer away is crucial to the ability to construct high performance structures from inexpensive feedstocks.

A. Herman, writing in [Modelling Simul. Mater. Sci. Eng., 7:43-58 1999] describes quantum mechanical simulation of mechanosynthetic steps useful in extending a silicon lattice. This paper is an extension of Dr. Herman’s earlier paper in [Nanotechnology, 8:132-144 Sep97].

The simulations were performed at room temperature, using the AM1 semi-empirical quantum mechanical model to calculate forces during the molecular dynamics. The reactions covered were insertions of :SiH₂ groups and isolated silicon atoms into a sequence of model compounds which covered all of the local environments found during the construction of a new layer on top of a hydrogenated Si(111) surface. In particular, addition of an Si atom to the most congested site needed for extension of the surface is simulated in this paper’s reaction 10. This reaction is equivalent to addition to a kink in the edge of a growing single-layer step.

Demonstrating that an addition can occur to this site and its diamond analog with an acceptable error rate in the presence of thermal noise is probably the most constraining local analysis in the path to atomically precise nanotechnology. The simulations in this paper show that adding to this site can be done in the presence of thermal noise and atomistic dynamics. The simulations reveal “the importance of initial SiH₂ or Si placement and confirms the occurrence of an optimal place for mechanical positioning of this reagent.” It would be helpful if further work explicitly described competing misreactions from placement errors and explicitly computed the magnitudes of the placement tolerances revealed by the simulations.

Dr. Herman favors silicon over carbon for initial mechanosynthetic work because it and germanium are “more readily subject to mechanochemical manipulation with [the] STM owing to their larger sizes and lower bond strengths and stiffnesses.”

The use of isolated silicon atoms as reagents in these simulations bring several considerations to mind. On the one hand, a truly isolated atom can’t be manipulated by forces on it from bonds from other atoms, so it seems a poor candidate for atomically precise operations. It can be manipulated with electromagnetic fields, but potential traps based on such fields normally have much broader minima than those due to covalent bonds. On the other hand, silicon atoms have been extracted from surfaces and deposited on them from tungsten tips in the AONO Atomcraft project. It would be helpful if the simulations were extended to explicitly model some STM tip atoms, determining whether Si atom interaction with the STM tip is compatible with the desired insertion reactions.

Dr. Herman proposes constructing atomically perfect STM tips as an application of early mechanochemistry. Given a commercial tip with a 5 nm radius, a sequence of 248 addition reactions is proposed to build up a consistent, atomically precise asperity on top of the bulk-fabricated tip. This asperity would contain 362 silicon atoms (including those from the commercial base) and would be 8 monolayers high. The paper notes that 0.1% error rates are needed for this application, providing another motive to calculate predicted tolerances explicitly.

New Components

New materials, or extension of our abilities to position old ones extend the set of usable candidates for initial molecular machine designs. The two papers summarized in this section extend the utility of ionic lattices and hydrogen-bonded supramolecular structures to nanometer scale fabrication.

J. Aizenberg, A.J. Black, and G.M. Whitesides, writing in [Nature 398:495-498 8Apr99] report controlling the crystal growth of calcite, CaCO₃, on a patterned substrate. They were able to control “…the location and density (N) of nucleating regions on the surface, the number (n) of crystals that nucleated within that region, and the crystallographic orientation of these crystals.”

They controlled crystal growth by depositing SAM (self-assembled monolayers) of polar thiol compounds in selected locations on metal films. Their polar thiols were HS(CH₂)_nX, with X = -CO₂H, -OH, and -SO₃H and n in the range from 11-22. The polar thiols were deposited by “microcontact printing with an elastomeric stamp,” which was itself patterned by casting silicone against a lithographically patterned master. After deposition of the polar thiol, the remainder of the metal film was passivated by deposition of a nonpolar SAM of HS(CH₂)₁₅CH₃. The substrate was then immersed in a supersaturated calcite solution. Crystals nucleated on the areas covered with polar thiols but not on the nonpolar thiols.

From the viewpoint of nanotechnology the most important feature of this work is the control of crystals’ orientation. By varying the polar moiety in the thiol and the metal film on which the SAM was grown, crystals could be grown with at least 4 different orientations with 98% accuracy, “implying that the atomic-level interfacial structure is controlling the nucleating plane of the crystals.”

As this technique stands, lateral control is much less precise. The crystal patterns reported have pitches ranging from 50 microns to 3 microns though the “method generates patterns of crystals with edge resolution of a few hundred nanometers.” Perhaps polar SAMs modified by AFM tips could create finer patterns.

The authors attribute the suppression of crystal growth on the nonpolar areas of their substrates to active depletion of calcite by growth on the polar SAMs. In a control experiment where only a single isolated dot of polar thiol was printed on an otherwise nonpolar surface, crystal growth was suppressed on the nonpolar surface within 80 microns of the dot, but occurred outside this radius.

The authors were able to tailor supersaturation, thiol and metal type, and pattern pitch to reliably nucleate one crystal per printed dot over a 100-fold range of dot densities. Presumably the absence of secondary crystals on these dots implies that crystal growth covers the whole dot before a second crystal nucleates. Calcite, like other ionic crystals, is a fairly stiff material (with a Young’s modulus of ~80 GPa). It is not the first choice for nanostructures, but it is stiffer than biopolymers, and its lattice is inherently a fused polycyclic structure. With finer lateral control, SAM-induced crystal growth might provide a route to constructing atomically precise parts from it or other ionic solids.

L.J.Prins et al. writing in [Nature 398:498-502 8Apr99] describe the synthesis of organic molecules that hydrogen-bond into chiral (lacking mirror symmetry) assemblies although only one of the types of components in each assembly is chiral. Their assemblies contain 9 covalently bound molecules apiece.

The particular class of structures that they synthesized consisted of 3 identical chiral pieces, each of which contains two melamines (-C(NH₂)=N-)₃ each with a chiral side chain, bound to a large central calixarene ring, and six molecules of a substituted barbituric acid. The structure consists of 2 sets of six-membered rings, each ring consisting of 3 melamines and 3 barbituric acids in alternating positions. Each adjacent pair of a melamine group and a barbituric acid molecule is held together by three parallel hydrogen bonds, much like the bonding within base pairs in DNA.

From the viewpoint of nanotechnology, these sort of self-assembled structures are useful as a mechanism for building larger atomically precise structures than purely covalent synthesis permits. Hydrogen bonds are “softer”, lower energy structures than covalent bonds, with lower activation energies for their formation. By the same token, they also form structures which are weaker and less stiff than covalent bonds form, so structures built from them are more distorted by thermal motion than structures which are purely covalent. Ideally, a self-assembled structure would contain a set of molecules, each one of which was different, each of which contained some optional groups which could be modified without damaging the self assembly process. Under those conditions the non-covalent bonds in the self-assembled structure would effectively gather together all of the design freedom of the constituent molecules into a larger structure.

In this work’s assembly, the lack of mirror symmetry adds a little design freedom. The melamine’s side chains can be a wider variety of moieties than symmetrical side chains would permit. The authors synthesized 8 assemblies with the same general structure. The assembly does, however, have an axis of 3-fold rotational symmetry and 3 axes of 2-fold rotational symmetry, forcing all of the barbituric acid molecules, for instance, to be identical. Ideally we would break all these symmetries and modify each molecule separately.

Single Molecule Techniques

The March 12th 1999 issue of Science had a special section on single molecule techniques. From the perspective of nanotechnology, the detection, characterization, and manipulation of individual molecules is close to the core of what we need.

Proximal probe techniques have been used to perform reactions on individual chemical bonds. If this route is used to construct nanoscale machinery, the initial generation of machines will be constructed one by one, essentially building individual hand-crafted molecules. The diagnostic techniques described in this issue, which can probe individual molecules, will be valuable in confirming construction steps, geometry, function, and dynamics.

These techniques may also be useful debugging self-assembled structures, though a larger range of existing alternatives are available for probing structures which can be produced in macroscopic quantities.

W.E. Moerner & M. Orrit, writing in [Science 283:1670-1676 12Mar99] summarize current techniques for the fluorescent detection of single molecules in liquids and solids. These techniques can probe the local environment of the fluorophore, sometimes with less perturbation than scanning probe techniques introduce.

Observing single molecules avoids averaging over an ensemble, allowing differences between molecules and changes in the behavior of the same molecule over time to be observed. For example, a fluorophore bound to an enzyme can display changes in fluorescence as the enzyme binds a substrate, triggers a reaction, and releases its products. In an ensemble measurement, detecting the equivalent changes requires synchronizing large numbers of molecules, which can be difficult, particularly for sequences of several states.

Two rather different regimes are room temperature studies of dye molecules and cryogenic studies of rigid polycyclic aromatic hydrocarbons. The dye molecules typically have absorption cross-section of ~0.06 nm² with an absorption spectrum 1000 cm^-1 wide. At cryogenic temperatures the spectrum of the aromatic molecules is a million times narrower, increasing their absorption cross-section to ~10⁴ nm².

The high cross-sections and narrow linewidths of the aromatics allows them to probe many features of their local environment. The resonance frequency of the molecule depends on local strain, for instance. Differences in resonant frequency can also be used to separate signals from molecules that are too closely spaced to be separated by light microscopy. Perhaps several probes with deliberately different resonances could be bound to a single nanomachine and monitor several degrees of freedom concurrently.

Room temperature dye molecules are useful in probing aqueous or other liquid phase systems. “Researchers have analyzed every aspect of the photons emitted, such as the[ir] location, polarization, time dependence, and spectral content.” Polarization could be particularly useful in probing machinery, because it informs us of the orientation of the fluorophore.

Another technique which yields geometrical information is fluorescence resonant energy transfer (FRET). In this technique a donor molecule absorbs energy from the exciting beam, but then transfers the energy to an acceptor rather than re-emitting it as fluorescence itself. The efficiency of this transfer drops as the separation between donor and acceptor is increased. FRET probes distances in the 5-9 nm range.

These techniques may be useful in probing the geometry and dynamics of nanomachinery, particularly in situations when scanning probes either perturb the machinery too much or cannot reach an embedded moving part.

M. Grandbois et al., writing in [Science 283:1727-1730 12Mar99], report experiments which directly measure the rupture strength of covalent bonds.

The authors attached polymer strands to an AFM tip and a substrate, then stretched the strand, measuring the forces as individual bonds snapped. The authors’ chose their attachment chemistry (succinimide side groups on their amylose polymer and amino bearing surfaces) to form multiple covalent links to each strand. Loops of varying sizes (0.6-60 nm) were present between binding sites. Bond rupture was detected by the temporary drop in force as loops unwound. The polymer backbone remains intact during these sequences of ruptures until the last rupture completely separates the tip from the substrate.

The forces at the start of each bond’s rupture were histogrammed. The histograms peaked at 2.0 nN ± 0.3 nN for amylose bound to silica and at 1.4 ± 0.3 nN for amylose bound to gold through a thiol.

The authors attribute the rupture strength on silica to Si-C bonds and on gold to either Au-S bonds or to Au-Au bonds. The assignment of these strengths relies on theoretical calculations of the strengths of bonds present in the attachment groups and in the backbone. The C-C and C-O bonds present in the backbone must not have broken since multiple detachments were seen in each experiment rather than only a single backbone cleavage. At the loading rate used in this experiment (10 nN/sec) these bonds and the C-N bond in the amide attachment are expected to break at ~4.1 nN while Si-O bonds should break at 3.4 nN and Si-C bonds at 2.8 nN. The authors attribute the 0.8 nN difference between theory and experiment to solvent effects.

The authors had to ensure that only single polymer strands were bound between their tips and their substrates. If multiple strands were bound, forces from several parallel bonds could be present, invalidating the data. They ensured this by measuring conformational changes in their strands at 0.275 nN, where amylose stretches by 0.05 nm per ring unit at constant force, producing a plateau in the force-extension curve. Parallel strands “either shift the plateau to higher forces or smear it out,” allowing the exclusion of data from these cases.

This experiment directly probes the thermomechanical cleavage of covalent bonds. It helps test the model of this mechanism in Nanosystems. This cleavage process is central to determining the lifetime of high performance diamondoid mechanical components under load. It would be helpful if the authors extended their experiments to test the temperature dependence of the rupture strengths they observed, which would further constrain the models.

This experiment is also important because it demonstrates, in the authors’ words, “the mechanical activation of chemical bonds.” Presumably the bond ruptures observed here generated reactive species such as free radicals. These species should be able to undergo reactions typical of diamondoid assembly proposals, such as addition to unsaturated feedstocks.

A.D. Mehta et al. writing in [Science 283:1689-1695 12Mar99] describe use of optical traps in measuring the properties of natural molecular motors.

A dielectric bead, typically silica or polystyrene spheres roughly 1 micron in diameter, is attracted to the focus of a laser beam. Although this location is stable, the forces from the trap are small, typically a fraction of a nN, giving spring constants of typically less than 0.001 N/m. These low spring constants are desirable in measuring the pN forces from individual molecular motors.

A typical experiment cited by the authors was Svoboda et al.‘s measurement of kinesin’s step size (8nm) and stall force (5 pN). In the cell, kinesin drags objects along microtubules, which are parts of the cellular cytoskeleton. Kinesin is powered by the hydrolysis of ATP. In Svoboda’s experiment kinesin was deposited on silica beads, allowed to attach to a microtubule (itself attached to a microscope coverslip), and its motion monitored via bead movement as the optical trap ramped up the load on the molecular motor.

The authors covered studies of:

• rotation of ATP synthase via imaging a bound actin filament
• kinesin movement along microtubules at varying ATP concentrations
• myosin pulling an actin filament
• RNA polymerase (RNAP) pulling itself along its DNA template
• force-extension curves of titin (a muscle protein with multiple folding domains)

All of the studies reported seem to be at the level of modelling the existence of a few major states in the working cycles of biological motors and the transitions between them. In the case of kinesin, for instance, it is still not known if the two heads move in a hand-over-hand gait over the microtubule or whether the working stroke of one head drags the other along at the same time. The authors do suggest that “Rigid links, defined protein orientations, and feedback systems should enable a host of single-molecule measurements that are analogous to seminal experiments with whole muscle fibers, observing conformational changes directly and perhaps arresting stable intermediates by applied load.”

In order to use natural molecular motors to help achieve positional control of molecular systems, I think we will need a great deal of information about the intermediates that occur in these motors’ working cycle. Ideally, we would like to single step these motors. This would require designing reversible inhibitors that prevent progress through at least two different stages of a working cycle. This will require detailed knowledge of the structure of the intermediates in a cycle.

Jeffrey Soreff is an IMM research associate.