Creating the First Omnifex

Back to Contents

    I originally composed this essay in Summer 1989 and I have updated it only slightly.


    The Omnifex provides us with a good example of Arthur C. Clarke’s dictum that "any sufficiently advanced technology is indistinguishable from magic". This Santa Claus machine awaits only a properly crafted spell entered into its computer to provide us with whatever we want. With assistance from a team of robots (some assembly is required) it can even create more Omnifices and Omniphages. Once established, the Omnifex technology will become self-perpetuating. The problem before us is that of getting it established. If it takes an Omnifex to make an Omnifex, how will we create the first Omnifex?

    A little simple arithmetic reveals the magnitude of the problem. If the Omnifex assemblers are spaced one hundred Ångstrom units apart, then the Omnifex plate will carry ten billion assemblers on each square millimeter of its surface. That means that your big Omnifex, with its one-meter by two-meters plate, carries about twenty billion trillions of assemblers. And although they are all identical, those assemblers are fairly complex arrangements of a large number of atoms. To make the problem even more complicated, the Omnifex requires structures in addition to the assemblers: it requires the channels and nodes that distribute atoms, wiring on the atomic scale to move electricity and data, and the capillary network with its atom grabbers as three obvious examples. There’s no way we could ever build something like that directly. We’re going to have to do something different, something that plays on the fact that the Omnifex is essentially a highly evolved descendant of the atomic-force microscope.

    (Speaking of highly evolved descendants reminds me of an interesting fact and spins it out into an amusing thought. When Bardeen, Brattain, and Shockley built their first crude transistor in 1948 they could not have imagined that twenty-seven years later microscopic versions of their invention, engraved and etched onto tiny chips of silicon, would be the basis for desktop personal computers far more powerful than any calculating machine then in existence or contemplated. The first crude atomic-force microscopes were built in 1985. Does that mean that we can expect to see the first Omnifices becoming available around the year 2012? We’ll see. [No, not even close: 2014 Mar 01].)

    The current first-generation AFM (as of 1989) is too crude to play any direct role in the creation of the Omnifex. It can, though, build a finer, multiheaded version of itself, a second-generation AFM, that will be used both to test Omnifex assembler designs and to build the third-generation AFM. Each succeeding generation of AFMs will be more complex and more powerful than the one that preceded it, so by the time we develop the fourth or fifth generation we will have a functioning, albeit small, Omnifex.

    I have no doubt that even now (as of 2010 Sep 20) some teenaged nerd has connected his or her computer to an atomic-force microscope (AFM) and has begun experimenting with assembling objects atom by atom. Wielding the AFM as Georges Seurat wielded his brush, our experimenter creates pointillist images from atoms on a suitable substrate. And just as Seurat had to return his brush to his palette to pick up a new dab of paint, so the AFM artist will have to return their atomic-force microscope to their "palette" to pick up another atom or molecule for placement. It’s crude and it’s slow, but it’s how we will get our second-generation AFM.

    Creation of the second generation of AFMs begins with the preparation of a substrate on which the assemblers will be erected. A square chip of some suitable material, such as silicon, perhaps a tenth of a millimeter long on each side, will provide a suitable substrate. Using the same techniques that are currently used to create computer microchips, the experimenters working on this project will lay down a network of electric circuits and atom-carrying channels leading from the edges of the chip to one hundred evenly spaced sites. Ending at the perimeter of each site, those wires and channels may be as much as a micron wide (and lest you think that’s a bit narrow, let me point out that sending atoms down a channel one micron wide is equivalent to sending basketballs down the Mississippi River). Thus prepared, the substrate will be mounted on a piezoelectric base, which in turn is mounted on an optical-quality motion-control platform, a combination that will enable the controlling computer to move the substrate in three dimensions with an accuracy of a fraction of an Ångstrom unit. At that point the substrate will be ready and will be moved into the volume of space accessible to the tools wielded by the construction computer.

    Using micromanipulators, the construction computer will first attach to the appropriate points on the edges of the chip the electrical leads and hair-thin tubes through which power and matter will be fed to the chip. Once that task is completed a supply of atoms and electricity can be fed to each of the assembler sites on the chip. The construction computer will then use a first-generation AFM to generate a map of the chip for use in guiding the construction of the assemblers. This step is comparable to creating a geological survey map of an area before building a railroad or highway through it. It’s going to be a big map: if instead of atoms we were using objects ranging in size from soccer balls to beach balls, then our project would involve building one hundred factories on a square two hundred miles on a side (and the channels would indeed be the size of the Mississippi River).

    Actual construction begins with the AFM grading the hundred sites, leveling them by moving the atoms that make them up. Then the construction computer will make the AFM reconfigure the channels and wires coming into each site by rearranging the atoms at their ends to make them narrow down to widths compatible with the assemblers. In the case of the channels that step will be comparable to narrowing the bed of the Mississippi River to the width of an irrigation ditch. Once that’s done the construction computer can begin feeding atoms up the channels to the sites and the AFM can use those atoms as building materials in the actual construction of the assemblers.

    Even with a supercomputer guiding the work the construction of the one hundred assemblers will take weeks, possibly months. There’s a lot of atoms to be moved and, as any experienced engineer knows, the project will encounter problems that will introduce delays. Channels will clog or break; wires will overheat and melt; electric current in one wire will interfere with current flowing in a neighboring wire; and so on. Each problem will bring its own delay and many will bring additional delays as the AFM is diverted from its construction task to repair damage or to rebuild some flawed element in the design. Eventually, though, one of the teams working on this project will solve all the problems their computer encounters and come up with a verified working second-generation AFM.

    Testing that new AFM involves hooking it up to its own supercomputer, by making the construction computer use its micromanipulators and its own AFM to make the connections at the edges of the chip on which the new AFM is built. Thus connected, the multi-AFM will be made to explore and map a variety of flat surfaces, mainly as a test of how well one hundred AFMs can be made to work in concert. The most important test will be that in which the multi-AFM creates a duplicate of itself.

    It will start with a bare chip of substrate material. None of the channels or wiring will be built onto the chip by the relatively crude photoengraving techniques of the computer makers. Instead, the multi-AFM will build all of the structures on the chip. In doing so, it will add features that it does not possess itself. Because it will be making the channels the correct size, only a few Ångstrom units wide, it will be able to lay down the correct number of channels, one for each of the chemical elements that will be used by the device, along with the nodes where the atoms are gathered and whence they are distributed to the assemblers. When that copy is complete it will constitute a true second-generation AFM and will be used to generate further copies of itself for distribution to all of the laboratories participating in the Omnifex Project.

    In addition to making copies of itself, the second-generation AFM will make working models of the various designs for the mechanism that will remove specific types of atom from the chemical soup that will be the Omnifex’s raw material. This may be the most difficult phase of the project. Chemists will be required to design fifty or sixty different kinds of molecular trap (and ultimately ninety-two or more) to grab specific kinds of atom from the chemical soup and push them into the appropriate channels. Each trap must be set by a small pulse of electricity and be sprung only by the right molecule. When sprung the trap must push the molecule into a second mechanism that strips any extraneous molecular pieces from the desired atom and shoves them back into the soup while pushing the desired atom into the channel that runs behind the trap. Though computer simulations will provide a fair indication of how well they will work, the designs proposed to carry out those tasks will have to be tested, both to allow unsuspected phenomena to become manifest for further study and to gain experience in the making and using of these broth-processing micromachines.

    The third generation of AFMs will be built by the second generation and will incorporate the improvements that the second generation was made to develop. Again the experimenters will start with a substrate made by other means. It will be a slab perhaps one millimeter wide in which hair-width capillaries and wires that enter through the bottom branch into submicron capillaries and wires that emerge through the top. Where they emerge at the bottom those capillaries and wires will have microscopic mating devices with which micromanipulators can connect them to sources of data, electric power, and chemical soup. The bottom of the slab will also carry the miniature posts that will hold it on its motion-control platform. Where the previous generation of AFMs received power and matter from the sides of its chips, the third generation will receive them from the bottom, as all Omnifices must do if the resources are to be distributed evenly to all parts of the Omnifex plate.

    Onto that substrate a second-generation AFM will build a network of wires, capillaries with their molecular traps and stripper gates, and atom-carrying channels, building them onto the wires and capillaries that emerge through the top of the substrate. Because the third-generation chip is larger than the second-generation chip, the second-generation AFM is going to have to build the third-generation prototype in stages, laying down a few layers of material on one section, then moving to a neighboring section and laying down the same layers, then moving on to the next section, and so on. To make the task more demanding the assemblers on the third-generation AFM will be packed more densely, spaced one micron apart rather than ten. As in the second generation, the assemblers themselves will be immobile, relying on gross motions of the chip to reach their assigned sites rather than relying on any motions of their own.

    This third-generation AFM will be an Omnifex of sorts, so part of its testing program will involve the making of millimeter-sized samples of various materials and objects. It might be instructed to lay down on a suitable substrate a sample of crystalline material, anything from simple sodium chloride (table salt) to a perovskite superconductor. Next it might be tested on the making of less regular materials, such as glass and plastics. It might be instructed to make a miniature junction-diode laser or a computer microchip. It might also be instructed to make samples of "impossible" materials, such as foamed steel or diamond shot through with threads of ruby. It will, of course, be used to make copies of itself.

    The fourth generation may be the prototype of the true Omnifex or the intermediate stage just prior to it, depending on how well the development is progressing. In either case it will be the stage in which the assemblers are made capable of the side-to-side motions that enable them to scan an area assigned to them. With that capability the fourth-generation AFMs and those of subsequent generations will be able to generate their products directly off their top surfaces: they won’t need a substrate to serve as a foundation on which to build the product, so they won’t need to be mounted on motion-control platforms. Instead, these AFMs will be mounted on solid bases through which they will be connected to their computers and their sources of power and chemical soup.

    The prototype of the true Omnifex, whether of the fourth or fifth generation, will be a plate perhaps a little over one centimeter long on each side and about one centimeter thick. Most of that thickness will be taken up by the multi-level branching that spreads the two- to three-millimeter power, data, and soup feeds entering the plate through its bottom out into the network of submicroscopically thin wires and capillaries that fills the lowest level of the AFM array. Such plates will be suitable for attachment to their bases with human-scale motions; that is, a human could, in theory, attach such an Omnifex plate to the rest of the Omnifex apparatus without breaking anything (in practice, though, robots will carry out the assembly).

    Once that prototype has been tested and is ready to make copies of itself the Omnifex Project will reach its flashpoint. It will have taken years to create this little one-square-centimeter-plus Omnifex but within a year tens of thousands of square kilometers of Omnifex plate will be in use, enough for all Humanity.

    The first product of that Omnifex miniplate will be a bar-like Omnifex plate, a plate a little over one centimeter wide and over one meter long. To produce it the miniplate will be turned on its side and the bar will come out onto rollers that will support its weight while it grows. If we assume that each assembler in the Omnifex makes one million atom-laying scans per second and that the average distance between successive atomic layers in the materials we want to produce is two Ångstrom units, then we may assume that objects will come out of the Omnifex at the rate of seventy-two centimeters (about two-and-a-half feet) per hour. At that rate our Omnifex bar will be ready in about an hour and a half. Actually the creation of an Omnifex plate will likely proceed at a slower pace because the complexity of the product will slow down the creation process (no shortcuts will be possible). But even if it takes a day or so to create the meter-long plate, representing a twenty-fold decrease in our assumed speed of creation, the results will still be spectacular.

    When that Omnifex bar is hooked up to its resource base it will be used to create an Omnifex plate somewhat longer than one meter on each side. Again the product will be created horizontally and pushed out onto supporting rollers lest its weight destroy the Omnifex creating it. That plate, in its turn, will be hooked up to a suitable resource base and will be used to create one hundred Omnifex plates one meter wide by two meters long (the plate for the standard workroom Omnifex). To grossly oversimplify what follows let’s assume that after creating those one hundred plates the Omnifex creates the parts needed to make the resource bases and attendant robots for four of them.

    Once assembled and activated, those four new Omnifices each create the means to finish up and activate four more and then those twenty Omnifices each create the means to finish up four more, thereby enabling the remaining eighty plates to be incorporated into functioning Omnifices. Robots work twenty-four hours a day with few breaks, so it should not take long for them to get all one hundred Omnifices up and running. Now we can repeat the process with each of those Omnifices. If the process takes one month, then at the end of the first month we will have 101 Omnifices operational; at the end of the second month we will have over 10,000 Omnifices up and running; at the end of the third month we will have more than one million Omnifices working for us; at the end of the fourth month we will have over one hundred million Omnifices, roughly enough for every household in America to have one; at the end of the fifth month we will have more than ten billion Omnifices, more than enough for every member of Humanity to have one; and at the end of the sixth month we will have over one trillion of these wonderful machines, more than one hundred for every human alive.

    Of course, the introduction of the Omnifex technology into human society won’t be quite that simple. In addition to the creation of the large number of Omnifices of several different kinds and the armies of robots that come with them, we will need a similar number of the various kinds of Omniphage, a whole new infrastructure for the distribution of power, data, and chemical soup, and more robots to put it all together and maintain it. The process will still be one of spectacular geometric growth, but it will probably take more like a year or two to complete. At the end of that time we will have achieved what could rightly be considered the apotheosis of Humanity: we will have acquired the power of gods and with it we will face the greatest challenge in our history.


Back to Contents