Institute of Atomic-Scale Engineering

The Shape of Things to Come
 By Forrest Bishop

Shape-Shifting Matter, Self-replicating Interstellar Nanoprobes, Drexler Universal Assemblers... Appeared in NanoTechnology Magazine - April 1996 Issue. Pepper T. Kim layout, Copyright (c) 1996 Forrest Bishop, All Rights Reserved.


 Our ability to imagine must always exceed our physical grasp; every conscious action is preceded by thought. The following is an overview of several interrelated ideas, not all of which may be realizable. If there is validity in the general concepts here, as well as in the diamondoid manu- facturing paradigm [1], we may see Shape-Shifting Terminators, self-replicating interstellar nanoprobes, Drexler Universal Assemblers, the terraformation of the Solar System, and more, within our lifetimes.

 The Shape Shifter

The Shape Shifter is an aggregate of "standard active cells", and other specialty cells of various dimensions and descriptions. An "active cell" is a space-filling polyhedral construct having power, signal, drive, and mechanical interfaces on each face of the cell [2]. A connected collection, or aggregate, of these cells form a "kinetic cellular automaton" capable of radically changing its shape. It is important to realize that a connected group of cells can be moved as a single unit some distance, then some new group can be formed and moved, perhaps in a different direction.

The size of the individual active cell is arbitrary, just as is the size of a pixel on a two dimensional monitor screen. The highest quality monitors use very many, very tiny, pixels to achieve their superior resolution. By analogy, the most useful and intriguing active cell would be very small, and atomically precise. A properly constructed "MNT Active Cell" should be able to move around without ever wearing out.

If you had a bread-box size piece of this "material", it might take the place of most of your electronic and household appliances by simply turning into the machine you wish to use at the time. Perhaps you would have it walk the dog (or simply be the dog), clean the house, and fix dinner in its free time. Tendrils of it might be used to perform microsurgery and other very delicate operations, such as removing plaque from an artery wall. As this is a machine that can radically alter its shape and surface composition, most applications haven't even been dreamed of.

 An MNT Active Cell

For the purpose of illustration, a particular "XY cube" MNT active cell was designed and characterized. Here 'XY' (and 'Z') refers to the local coordinate system of the particular face of the cube under consideration, not to any global coordinates. "XY" refers to the fact that the cells have mechanical interfaces on each of their six faces, that allow sliding two interfaced cubes in one of the two directions 'X' or 'Y'. The cubes are unable to detach the two joined faces by movement normal to the plane of the joint. They can only move in one of the two permitted directions at a time, and must be aligned with four faces flush in order to change direction. .It has been shown [2] that the restriction to only sliding does not incur significant penalties on the possible movements of the cells, while providing a number of advantages.

 This baseline design has a nominal edge length, or cell metric, of 167 nanometers. The principle reason for the selection of the cell metric was to adhere to the design rules in [1], especially regarding size and spacing of conductors. It should not represent a lower bound on the size of an active cell. Secondary constraints include structural considerations, as well as the size of the included motor and controller.

 The structural material is of the diamondoid class, assembled by the putative methods of molecular manufacturing [1]. Each active cell contains an internal Drexler "rod logic" [1] controller to perform various housekeeping functions, as well as to communicate with its neighboring cells.

 When two XY cubes are interfaced and aligned, some method of locking them together is desirable. This might be done piezoelectrically, but the baseline design uses four tapered, retractable locking pins extending from one cell to complementary holes in another. There are other mechanical methods of accomplishing the same thing.

 There is, in addition to the controller, an internal electromechanical interface switch for power and signals. Energy is delivered to the cell electrically, via roller contacts. The drive system looked at in some detail is a linear electrostatic motor derivative. The conductor material is unfortunately not specified. As the electrical lead and contact resistances cannot be fully characterized, a programmable voltage multiplier is incorporated to keep the cell-to-cell energy transfer at the nominal voltage. This imposes a major signal propagation delay, which might be decreased, even to the theoretical limit (with superconductors), in a more refined design.

 It should be noted that a system composed of these kinds of active cells may form the core of (in addition to a Universal Assembler) a research facility inquiring into many aspects of atomic-scale engineering. Some of the research would, in turn, produce data for better design of the active cell.

Geometrical, Mechanical:

The mechanical interfaces consist of orthogonal "T-slots" cut into a face of a cube. The slots are parallel to the edge of the face, and actually form "T-posts" when both sets of slots are considered. There are two complementary sets of these T-posts, and each cube has one type ("active faceplate") on three of its adjoining faces, and the other type ("passive faceplate") on its other three adjoining faces. Although the number of T-posts is somewhat arbitrary, the reference design uses nine T-posts on the active face, and 16 T-posts on the passive face.

These features of an XY cube introduce a chirality to the aggregate, and a specific orientation for mutual interfacing. As there are no permitted modes of rotation, this internal orientation is maintained regardless of the configuration of the aggregate. An active face is therefore always adjoining a passive face.

In order to move in the 'X' or 'Y' direction, the two pins on the axis perpendicular to the desired movement are first withdrawn. With the drive engaged, the two pins in line with the movement, which have tapers in this direction, are then withdrawn at some controlled rate.. Owing to this taper, the cell begins its movement before the pins are fully withdrawn. The purpose of all this activity is to prevent the cell from wedging by rotating about the local 'Z' axis. Once the cell moves out of the aligned rest position, this is no longer a major issue.

With the electrostatic drive systems outlined below, an active cell can be built that has no breaches, or holes, through the cell wall, an advantage in environments such as air. This particular design is composed of six individual faceplates (three of each type) which are then assembled, along with the internal parts, to form the active cell. The inclusion of these joints, along with the mechanically operated locking pins may limit their use to in vacuo.

Each of the active and passive faceplates has the same kind of edge treatment, consisting of a finger joint and other standardized features. As the joint surfaces are parallel to the cube faces, the faceplates can be assembled in any order. A pin introduced at one end of the formed edge joint holds the two faceplates together. The pins are always inserted through a hole in an active faceplate. In addition to simplifying the manufacture, this modularity physically allows a modicum of self repair (remove and replace), in the case of a radiation damaged space probe, for example. The "Overtool" concept also makes use of this feature.

Drive Systems:

It is possible to use rotating motors, clutches, pinions, racks on the faces, and so forth to drive an XY cube. The T-posts on the active faceplate might house the drive pinions and their associated bearings. The toothed racks are then incorporated on the channel surfaces of the passive faceplate. These pinions might also serve as rolling electrical contacts.

There are several alternative linear electric motor types of interest for the mesoscopic active cell. One is a linear version of the electrostatic motor presented in [1], Section 11.7. The tunneling contacts and variable work function surfaces would be on the faces of the active T-posts, with the conductors embedded in the channels of the passive faceplate.

This particular design uses the "Dielectric Drive". Consider two parallel, charged conductive plates as in a capacitor. If a slab of dielectric material, thin enough to fit in the gap between the plates, is introduced at one edge of the plates, it will experience a force tending to draw it into the gap.

When the dielectric slab reaches the far edge of a charged plate, no further motive force is available. It is then necessary to switch to another set of plates further along the path of motion. This is not unlike the switching in an ordinary forced-commutation electromagnetic motor.

 To implement this drive system, the undersides of each of the nine active T-posts are divided into several conductive regions, separated by insulating gaps. The "channel surface" of the active faceplate contains the ground planes, which together with the conducting plates embedded in the undersides of the T-post shelves forms the powered portion of the dielectric drive system. The passive faceplate is a solid piece of dielectric material (diamondoid), with interfacing conductors embedded in its channel surface.

The operating voltage range for the dielectric motor is quite variable, from greater than zero, to some value below the dielectric strength of the specified 3nm diamondoid insulating gap.For this particular design, at one volt applied, this yields an acceleration of about two million gravities, neglecting damping.

 Note that the cell being moved relative to the power supply isn't necessarily the one providing the motive power. For this and other reasons, it is desirable to utilize sliding electrical contacts in this design. This is another area where more research is needed.

Power and Signal:

The four locking pins also serve as the cell-to-cell electrical contacts when they are extended into their matching holes. They are spring- loaded so as to be extended when no power is applied (i.e. "deadman" switches). A lever from the rod logic engine operates the pins. Two of these pins form the sliding contacts when they are in their partially retracted position. The other two are not in use when the cell is moving.

A fifth pin at the center of the active faceplate is for a more direct and lower resistance power transmission path. More such pins and holes can be added if a finer step resolution or greater shear strength is needed. Note that the dielectric drive is capable of stopping in many places, depending on the layout of the conductive plates.

The power and signal interface switch is located at the center of the cell. A set of power and signal busbars from each face terminate in a mechanically operated, conductive sleeve (or roller contact) that can be extended to contact a conducting "routing" cube.

A serial signal bus may be implemented by modulating the power, but the voltage boosters add some complications to this. An alternative is a separate line for signal . Another alternative is acoustic transmission via the locking pins. These pins should include contact "switches", levers actually, to verify the locked and unlocked conditions. These same levers could be operated directly by the logic engine for serial signaling.

Research Facility

The ability to manipulate several very small objects in three dimensions is very limited at present. It may be possible to build active cells with current silicon micromachining techniques, perhaps combined with some "bottom up" components. These could be very useful in studying the behavior of materials at the mesoscopic scale. Any face of a cell can be replaced with some other tool besides the standard faceplate. It may be possible to use these microtech cells to construct an even smaller family of active cells. An intriguing variant would use larger active faceplates, connected to an external computer, to manipulate smaller passive faceplates not burdened with onboard microtech processors.

Space Probes

A Shape Shifter makes for a very interesting space probe. The ability to act as the structural material, computer system, attitude control, and reaction mass for the probe's various instruments and propulsion systems significantly decreases the amount of mass required. The Starseed/Launcher electrostatic accelerator proposal is built around these notions.

The Overtool

This is a proposed method of constructing a Drexler Universal Assembler. The essential features include the MNT Active Cell, described above, and a manipulator mounted on an active cell. An aggregate of these devices should provide the necessary functionality, variability, and means of transport to form the core of a general assembler, for macroscopic assembly as well as for diamondoid mechanosynthesis.

The active cells and manipulators are further broken down into a set of standardized parts. The assembler should be capable of making and assembling all of these parts, and thereby replicating itself.

The essence of this proposal is to reduce the number of individual parts (the active cells) to a small set of identical components governed by a few simple rules of interaction. A collection, or aggregate, of these cells then form a device of arbitrary size which can change its configuration to fit the desired task.

 A second type of cell has much the same functionality as the "standard cell", except that one of the faces of the cube is replaced by an "XYZ gantry", similar to a contemporary Coordinate Measuring Machine, to permit the fine control and generate the forces needed for diamondoid mechanosynthesis. This simple type of three-dimensional manipulator may seem to be too restricted in its motion to perform the many necessary maneuvers. However, when incorporated in an active cell aggregate, and programmed to work in concert with others of its kind, this machine can indeed execute a large class of rotational and translational movements of interest to the mechanosynthesist.

The XYZ Gantry has a base configuration that allows mating to the open side of the otherwise standard active cell. There are six possible gross orientations for this gantry (+-('x','y','z')). Additionally, since the faceplate joints are four-fold symmetrical, two or four 'XY' orientations are possible, depending on the symmetry of the gantry itself.

The tip of this mechanical arm may have a holder for interchangeable tools, or perhaps a dedicated tool. As there are six faces on a cube, there are six possible orientations for this "Gantry Cell". This means a number of Gantry Cells can be simultaneously engaged in a particular process. Some of the cells can be holding, straining, and rotating an arbitrary workpiece, while others fetch reactive species or perform abstraction reactions.

Although it is not necessary to use the XYZ gantry here, a number of advantages incur. Its geometry is quite compatible with the cubicle active cell, simplifying the design. In addition, the Cartesian kinematics are trivial, for both forward and backward solutions. This means the development time, as well as the real-time computation, are significantly less than what would be required for a more elaborate robot arm.

It may be possible to allow a portion of the 'Z' axis actuator to extend into the included volume of the active cell, depending on the arrangement of the internal parts of the five-faced standard cell. This is not strictly necessary. A Gantry Cell can be two or more times the standard cell edge length in its local '-Z' direction without putting undue constraints on its versatility. The gantry cell of figure (Gantry Cell) is designed to fit within a volume of two standard cells without interfering with the interior components of its active cell base.

A secondary feature is the "Moiety Palette", which is based on an active cell "passive faceplate". What would be the interior surface of this component is made flat, with receptors incorporated for the desired molecules. A similar device is used for waste removal.

With the above specifications, as many as 14 gantry cells of the six possible gross orientations can be brought together such that their working tips are confined to the same small volume (say 1000 nm^3, depending on the tip geometry). For example, four each of +-'z', two each of +-'y', and one each of +-'x', are brought together as depicted in the "14 gantry cells" figure. One of the 'z' cells is shown with a different 'XY' orientation than the other three, as mentioned above.

This configuration, or some part of it, would be useful in straining a bearing sleeve such as the one depicted in [1], Figure 1.1. The +-'x' tips are brought together to form a mandrel. Some portion of the remaining 12 available gantry cells (say, some of the four +'z') then bring a pre-assembled flat strip up to the mandrel and wrap (and strain) the strip around it. The four +-'y' cells can assist in this. While these cells hold the strip in place, another available gantry cell(s) (the four -'z') performs the necessary abstractions and reactions needed to finish the joining. The +-'x' tips then retract, freeing the finished part.

The architecture of an active cell aggregate is eminently scaleable, from the assembly of mesoscopic systems, to the mechanosynthesis and assembly of quite large structures using the same active cell family. The diameter of the shell in the "Strained Shell" figure is intentionally omitted. It may be 1700 nanometers, or perhaps 1700 kilometers (space-based). It would be necessary to build such a large structure of something very strong and cheap, like diamond.

Using the "Strained Shell" figure as an example, consider now that the active cells are not being used to strain the structure, but rather to transport gantry cells and moiety palettes to and from the entire exterior surface of the (unstrained) shell. This shell is now being constructed as a single unit, in a manner akin to crystal growth. This principle can be extended to more elaborate morphologies, as well as being combined with the traditional convergent assembly processes.


Our machines will come to resemble biological systems in their complexity, adaptability and agility. The above is only one genus. It is instructive to cast these directed, replicating machines in the light of a new form of intelligent life.

[Home] [Articles] [Contact] [Images]
Copyright ©1967-2004, Forrest Bishop, All Rights Reserved