MM primitive systems
From Wise Nano
Primitive Molecular Manufacturing Systems
Once a general-purpose manufacturing system capable of building duplicate and improved systems is developed, further progress and improvement can be rapid. However, it has not been clear what the first such system would look like, or how it could be built. This section proposes two architectures for primitive systems that could be developed with today's technology. Both use relatively uncomplicated nanomechanical systems, attached to larger actuators, to bind prefabricated molecular “nanoblocks” from solution and fasten them to chosen locations on the product. Each design is intended to be capable of manufacturing an array of independently controlled duplicate systems.
This section provides general background, then describes the physical layout and general functionality of each approach. Then it focuses on handling and joining nanoblocks and the functionality of nanoblocks and nanosystems. Finally it briefly analyzes the productivity of the general nanoblock-placement approach, and discusses how productivity could be scaled up.
Contents |
Background
It is too early to tell whether the first molecular manufacturing systems will be based on solvent-immersed mechanisms assembling prefabricated molecular building blocks or on scanning probe systems doing machine-phase mechanosynthesis to build covalent solids. As nanoscale technologies and molecular manufacturing theory have developed, it has at times appeared that the most effective development approach would be to develop a scanning-probe system that could do machine-phase mechanosynthesis, and use it to construct, slowly but directly, a nanoscale system capable of using machine-phase mechanosynthesis to build duplicate nanoscale systems more rapidly.15 Recent progress in molecular building blocks, along with more detailed understanding of how a primitive “wet” system could be improved incrementally to an advanced “dry” system, suggests that a “wet” context is also a good place to start. (It is worth noting that Drexler, who is often associated with “dry,” highly advanced diamond systems, has always recommended starting with “wet” systems.)
Several different classes of molecules can implement an engineered structure. DNA forms predictable folded structures, and a variety of shapes have been made with it. Although double-stranded DNA is too compliant to make multi-nanometer structural components, stiffer four-stranded variants have been used. Nadrian Seeman has made crystals up to a millimeter in size, ordered to 1 nm resolution.16 Protein design (which is easier than predicting how natural proteins will fold) has resulted in novel folds, and can be used to produce small complex shapes. Perhaps the most promising molecular backbone is being developed by Christian Schafmeister: a polymer that is relatively stiff even in single strands, using spiro linkages between cyclic monomers. A library of monomers allows the design of a wide variety of twists and shapes.17
To form a solid product, blocks must fit together. See Fig. 3. To help with alignment and insertion, a completed layer of blocks should form shallow pits between the blocks into which the next layer of blocks will fit. A cube standing on one corner would have an appropriate shape and orientation to form a solid structure, but many shapes will work. In effect, each completed layer forms a regular grid of vacancies for the next layer of blocks to be placed into.
Fig. 3. Tattooing a Row: Depositing nanoblocks via scanning probe to build an array of “Tattoo” block-placement machines. Note the orientation of the deposited blocks. The free-floating nanoblock is about the enter the corner-shaped cavity in the probe tip. The tip will then push the block into the nearly completed nanoprobe directly underneath it. When in operation, each nanoprobe in the row can be extended by establishing a voltage difference between the appropriate “x-wire” (the row of white conducting blocks that runs almost horizontally near the bottom) and “y-wire” (the conducting rows that extend toward the upper left).
A 5-nm molecular building block could contain thousands of atoms. This is small enough to build via advanced synthetic chemistry (perhaps aided by self-assembly), and to maintain a well-defined structure even if the framework were somewhat sparse, but large enough to include several molecular actuators, a molecular electronics component, a fullerene-based structure, or a fluorescent nanoparticle or molecule. The surface of the block would have room for several dozen amino acids or photochemical sites, the position and orientation of which would determine its bonding behavior.
“Tattoo” architecture
The Tattoo architecture for programmable heterogeneous assembly of nanoblocks is based on the possibility of making nanoblocks that will not aggregate or bond in solution, but will bond when pushed together by mechanical force. See below in “Handling and joining nanoblocks” for discussion of several ways in which this might be accomplished.
Given a way to reversibly bind a nanoblock to a weakly binding site (“socket”) attached to a scanning probe microscope tip, and a surface that (like the surfaces of blocks) will bond to a block that is pushed against it but will not accrete blocks randomly, the scanning probe tip can be used to deposit multiple blocks at selected locations on the surface. The tip with bound block might be used to image the surface, though fluid immersion (for non-contact or tapping-mode AFM), block fragility or unwanted bonding (for contact-mode AFM), and block conductivity (for STM) may present practical problems. Once a layer of blocks is completed, a second layer can be deposited on it, and so on. Solutions of different nanoblocks can be flushed in, allowing each layer to be made of as many block types as desired.
Once the ability to build patterns of blocks is established, the next step is to build a "tattoo needle." This is a socket attached to a nanoscale actuator which can be individually activated—via electrical connection, if the actuator is electrical (optical actuators will not need physical connection). The actuator needs only one degree of freedom. Its purpose is to push the socket away from the surface with a displacement on the order of a nanometer. (The "needles" will be positioned near the product by a large positioner after blocks have bound to their sockets, and selected actuators will be activated to drive the blocks the final distance into the product.)
Construction of the socket will probably require special design. Several blocks placed in a triangle will make a block-shaped cavity. In operation, the inner cavity must attract blocks but not bond to them; this can be accomplished by a charge on the inner faces opposite to the normal block charge. However, in solution, the cavity blocks must not aggregate with the product or with each other. Because multiple block types can be used, blocks with a single special face can be placed in a cluster to form the socket. A chemical post-processing step may be required to modify the special surfaces of the blocks. See Fig. 4A. Alternatively, a special prefabricated socket structure could be attached to a distinctive mounting point by self-assembly. See Fig. 4B.
A 
B 
Fig. 4. Two ways to implement a socket: A) Place several blocks to form a cavity, perhaps followed by post-processing to modify the inner faces. B) Use a special-purpose molecular building block.
Once the actuator-and-socket machine is built, it can be used to deposit blocks on a surface held in front of it. If that target surface is attached to a three degree of freedom manipulator, which is used to reposition the surface relative to the deposition machine, then blocks can be deposited in selected locations and in layers. (If a scanning probe microscope is used as the manipulator for the target surface, then scanning the surface relative to the machine's socket might be used to image the surface, though there are several practical problems with this.) The goal is that the machine should be able to build a second, identical machine on the surface that is presented to it by the manipulator; and not only one, but a row or array of machines. These machines could then be used in parallel for higher-throughput manufacturing of larger arrays.
If the machines can be independently controlled so that only some of them are made to deposit blocks at any given placement, then they can be used to build heterogeneous structures or regular structures with different periods than the machine spacing. Independent control can also be used for error correction; if a machine is known to be non-functional, another machine can be translated over its position to place the required blocks.
If different types of nanoblock-specific sockets can be built on different machines in the grid, then multiple nanoblocks can be mixed in one feedstock solution and each machine activated only when its selected type of nanoblock is wanted. This would remove the need to flush multiple solutions past the machines, eliminating the corresponding penalty in time, materials, and possibly errors resulting from contamination with the wrong type of block.
An even simpler variant of this approach might be useful to test the block-binding surface functionality before complicated nanoblocks and nanoscale machines are developed. A tower or needle built without an actuator but with a socket at the tip could be used to deposit passive chromophore-containing blocks by moving the surface it is attached to in a way that presses the socket against an opposing surface. The light from a simple heap of blocks would be visible with an optical microscope; several heaps spaced several microns apart in an identifiable pattern would confirm success. A more ambitious goal would be to use the tower to construct more socket-tipped towers, and then test their functionality. With sufficient precision, an exponentially growing number of towers could be created.
“Silkscreen” architecture
The Silkscreen architecture is based on the idea of separating a solution containing nanoblocks from the substance or condition that would cause the blocks to bond together. Instead of the array of "needles" in the Tattoo architecture, the Silkscreen is a membrane with an array of holes. The membrane serves several purposes. It separates the feedstock blocks from the product, and can maintain distinct conditions (such as concentrations of zinc) on each side. Its primary purpose is to control the position and timing of block passage through the membrane to the product.
Each hole in the membrane contains an actuator which can reversibly bind to a single block, transport it through the membrane, block the hole to prevent mixing of solutions, and present the block to the product. (Something shaped like a cutaway wheel or disc, with a socket in its rim and mounted on a torsion hinge, can perform all these functions simultaneously with only one degree of freedom and no bearings. See Fig. 5.) Like the Tattoo approach, the Silkscreen approach repositions the placement machine relative to the product using a three degree of freedom manipulator.
The membrane would be closely fitted to the growing product, and could be aligned to it by local forces. A block passing through a hole in the membrane must be able to reach only one vacancy in the product; the block's motion can be constrained by the manipulator until it is bound to the product.
An initial Silkscreen membrane might be built by any convenient combination of self-assembly, synthetic chemistry, and lithography. The grid of holes could be created either by lithography or by a self-assembled membrane such as DNA tiles. Each hole would be filled by a molecular actuator system. Once constructed, the first system could be used to build larger membranes and improved designs.
The simplest membrane might have only one hole. Its actuator could be activated by light (which would not need to be focused) or by passing current between the product and feedstock side of the membrane. The hole might be constructed by slow lithographic techniques such as ion milling or dip-pen nanolithography (DPN). Even with only one hole, the size of the product would be limited only by the speed and reliability of deposition and by the range of the product positioner.
A 
B 
C 
D 
Fig. 5. A) The manipulator is in position to gather a block from solution. B) A close-up of the socket's position with a block inserted. C) The manipulator has moved to attach the block to the product. D) A close-up of the block about to be pressed into place.
After manufacture of a membrane, passive gaps and actuators that are jammed open could be sealed shut by (for example) putting half of a binary adhesive on each side of the membrane. A plug will form wherever a gap allows the two components to mix.
Handling and joining nanoblocks
The feedstock of a primitive “wet” molecular manufacturing system will be prefabricated molecular or macromolecular blocks, a few nanometers in size, dissolved or suspended in a solvent. The function of the manufacturing mechanism will be to take individual blocks from solution, move them to a specified location in the product, and attach them strongly enough that they do not unbind. The design of the system must ensure that blocks in solution very seldom aggregate with each other or attach to the product where they are not desired, but once attached and bonded to the product they very seldom detach.
A charged object in solution will be surrounded by counterions of the opposite charge. In polar solvents like water, the solvent molecules will align to nearby charges and partially screen them; in nonpolar solvents, charges can affect each other over longer distances and are apt to bind to the object. If two objects of the same charge approach each other, they will repel. Less polar solvents will extend the repulsion zone, requiring more energy to force the objects together.
There are many ways that blocks can be strongly joined—a wide range of chemical reactions or intermolecular forces can be used. Two attractive possibilities are photochemical crosslinking and the binding of zinc or other metal ions to certain amino acids, both described in the Mechanosynthesis subsection.
For the purpose of joining blocks, the reaction should happen only when the blocks are pushed together, not when they are floating in the feedstock solution. If zinc binding is used, then excluding zinc from the feedstock solution and excluding stray blocks from the product area can prevent the feedstock blocks from aggregating in solution. The “Silkscreen” approach was designed to maintain a product environment of different composition than the feedstock solution. If photochemistry is used, then stray blocks must be prevented from contact while the photons are being delivered. Putting a charge on the blocks can keep them separated in solution, preventing unwanted aggregation even if no membrane is used to separate feedstock from product environment, as in the "Tattoo" approach.
A simple approach to block fastening is to cover the block with zinc-binding amino acids forming half-binding sites (two acids forming half of a tetrahedral site). Design the block with a negative charge equal to the number of sites. Zinc ions, with two positive charges apiece, will bind to each site, giving the block a net positive charge; this will keep it separated from other blocks in solution. If two blocks are pushed together strongly enough to overcome their repulsion, then half of the zinc will be squeezed out, leaving exactly enough zinc to neutralize the charge on the two blocks, and fastening the blocks strongly together. If the blocks are reasonably rigid, then it will be energetically unfavorable for extra zinc (along with its associated counterions) to squeeze in and allow them to separate, because multiple zinc ions would have to enter the tight inter-block space simultaneously. Thus the feedstock solution, with loose blocks, zinc, and counterions, could be allowed to contact the product without dissolving it or accreting to it. This would be ideal for the “Tattoo” approach.
Attachment mediated by photochemistry or electricity should work with either the Tattoo or the Silkscreen approach. Although it is somewhat more complicated, requiring delivery of light or electricity as well as some way to cancel the charge on the blocks as they are added to the product, it has the advantage that it will work equally well in more advanced solvent-free systems. Another complication is the need to keep reactive molecules (such as oxygen) away from the photochemical sites.
If the zinc binding approach and the photochemical approach conflict with some aspect of block or system design (for example, if the blocks cannot be prevented from colliding and accreting too frequently due to Brownian motion), it should be possible to use a pair of molecules that forms a bond via a reaction that is energetically favorable but has a high energy barrier. Such a reaction will happen very seldom between blocks in solution, because both the reaction's energy barrier and the block's repulsion must be overcome. But once blocks are confined and pressed together, the block's repulsion will no longer impose an energy barrier, and the effective concentration of the reactants will increase by several orders of magnitude; proper alignment of the reactants may also help. Together, these factors should make the reaction happen many orders of magnitude faster, allowing a fast assembly rate to coexist with a low error rate. Many such reactions will work without solvent.
In order to manipulate blocks mechanically, they must be attracted from solution and attached to a manipulator. This will happen if a binding site (“socket”) in the manipulator is made attractive to the blocks, for example by giving its interior a charge opposite to the charge on the blocks. This is an application of self-assembly. For some blocks, it will be important to orient them correctly. Patterns of charge, asymmetrical shape, short strands of DNA, and weak bonds such as hydrogen bonds can be used to cause the block to favor a particular orientation and can make a socket specific for a particular block type.
The manufacturing system will need to be able to place more than one kind of block. There are two ways to accomplish this. One possibility is that the block types will be mixed in the feedstock solution, in which case the sockets must be block-specific (meaning that they must be reconfigurable, or there must be multiple sockets). The other possibility is to flush through one feedstock solution at a time, with each solution containing a different block type. When a solution is flushed out, blocks will remain in sockets, but can be dealt with simply by depositing them onto the product in an appropriate location.
If the system uses zinc binding, then putting a few zinc binding sites in the binding surface of the socket can be used to bind to the block strongly enough to hold it reliably, but weakly enough to let it go without damage when the actuator is retracted from the product. Slight mis-alignment of the binding sites can reduce the binding force, and adding more sites can increase the force.
Nanoblock and nanosystem functionality
The simplest manufacturing systems only need to extend or retract a bound block; this requires only a linear actuator with one degree of freedom and a small range of motion, and a way to control individual actuators. There are several kinds of molecular actuators that may be suitable, and several kinds that are less suitable for one reason or another.
Speed and addressability will be important for any practical nanofactory. DNA binding actuation, though quite flexible, is also quite slow: many seconds are required for the strands to diffuse and link. Molecular precision and small size are important; this may rule out some actuators that depend on bulk effects.
Some molecules are responsive to light; they would have to be placed at least several hundred nanometers apart to be individually addressable, but this may be acceptable in early designs. Light has the advantage that it requires no physical structure to deliver it; it can be focused in a pattern from a distance. Also, light can be switched with very high bandwidth, though the response time of a slow actuator might negate much of this advantage. A focused pattern of light will have low spatial precision by nanoscale standards; this may be partially overcome if molecules can be found that are sensitive to specific wavelengths of light, so that several different actuators can be used within a single pixel of a multicolored pattern.
There are several kinds of electrically actuated materials. Piezoelectric materials deform because the spacing between charged atoms in the crystal varies slightly under an electric field. Although they require high voltages at millimeter scale, sub-micron thicknesses should be activated by sub-volt fields. However, they have very low strain (at most a few percent) so are probably not suitable. Some electrically deformable polymers work by ion intercalation/expulsion, which may release unwanted ions into solution and may not work in single molecules at all. Redox reactions can cause changes in a molecule's electron distribution, which can cause large changes in its shape or in how it fits together with other molecules. The mechanism of known redox actuators involves protonation, and this may not work in some environments; search is ongoing for molecular actuators that use only electron exchange. Annulene-based actuators can deliver strains of almost 6%.18 Poly(calix[4]arene-bithiophene) has been calculated to produce 78% strain, and another thiophene, poly(quaterthiophene) (poly(QT)), has exhibited reversible strains of 20% at 5 volts.19 Such actuators would seem to fulfill the required function. These redox powered actuators could presumably be protonated by battery-type reactions that are driven by varying electric currents from nearby nano-wires. A separate wire need not be run to every individual actuator. Even without digital logic, an X-Y grid of wires can be used to control an actuation system at each of the points where the wires cross.
Sensing will be important, not just to prevent accumulation of errors from generation to generation, but also for research: to provide early confirmation that new designs are working as intended. The important question will be whether a block has been placed as intended. The block can be detected by contacting it with a physical probe, which would be similar to a block-placement probe without the socket and with a weak actuator. Full extension of the probe would indicate an absent block.
Information must be returned from the nanoscale. Electrical signal return could use a simple mechanical switch, as suggested for digital logic. Information could also be returned optically; fluorescent nanoparticles can be held near quenchers when the probe is retracted, and extension of the probe would cause significant increase in fluorescence. Single-molecule fluorescence has been detected. If probes can be individually activated, then they can be spaced closer than the diffraction limit; probes spaced far enough apart could still be operated in parallel. Alternatively, careful detection of light levels could indicate how many probes within a single pixel had been extended, and this could be compared with the intended result; if an error was found, then single-probe actuation could be used to isolate it.
Throughput and scaleup of nanoblock placement systems
Activating a molecular actuator via photons or electric fields might take on the order of a millisecond. Diffusion of nanoblocks to fill the sockets might take several milliseconds; this depends on many factors, including block size, the concentration of the blocks, and the viscosity of the solvent. If ten five-nanometer blocks per second can be placed, then a single placement machine would be able to deposit a square micron area of a single layer, five nanometers thick, in a bit over an hour, or a 100 nanometer cube in 20 minutes. The placement machine itself will probably be built on a 100-nanometer scale (20 blocks on a side), implying a very high manufacturing throughput relative to its size.
Diffusion and binding of blocks to sockets is probabilistic, and without the ability to detect when a socket is filled, a relatively long time must be spent waiting until it is very likely to be filled. If this is the limiting factor in deposition speed, the ability to sense when a socket is filled would allow faster deposition. Operations could be scheduled as blocks became available. This might allow an average of 100 or even 1000 blocks to be placed per second per machine. A system containing 10,000 machines placing 1000 five-nanometer blocks per second apiece would build only a few nanograms of product per hour, a cube several microns on a side, but that corresponds to several billion blocks—enough to build a powerful CPU, for example.
