Suppose we have positional control precise to within a few microns, but not much force, power, rigidity, money, or access to pure materials. How can we leverage this into a comprehensive flexible digital fabrication capability?
The standard ghetto metalworking processes are stick welding, bending, drilling, hammering, and angle grinding.
Digitally-controlled stick welding is maybe a bit difficult, but should be feasible, since “welding robots” have been a thing for decades; controlling the power supply should make it much easier to strike the arc with a TIG-like high-frequency start, you can avoid sticking the electrode because you don’t need to enable high current until an impedance measurement shows the electrode has been lifted from the surface; voltage feedback should provide a trustworthy and low-latency indication of the arc length once it’s struck; current and polarity duty cycle can be adjusted rapidly to control electrode melt-off rate; and precise toolpath control should dramatically improve weld quality. To the extent that you can substitute a CO2 hose (“MAG welding”) for standard stick-welding flux, you can avoid slag that needs to be chipped or ground off; this should be no problem when welding on steel. “Short-circuiting metal transfer” as in MIG/MAG welding should be a possibility. This kind of process might allow “stick welding” using plain steel baling wire, and thus 3-D printing in mild steel.
Model-predictive control of the temperature distribution across the workpiece, with webcam feedback from surface temperatures, is probably critical to this. The dimensional precision of the result will probably be compromised by the contraction of the deposited metal, which is one reason welding is conventionally used for permanent assembly of prefabricated parts rather than large material buildups (though adequate toolpath planning could potentially reduce this problem), and the resolution will be limited by the surface tension of the weld puddle.
Bending of wire, rebar, or sheet-metal strip can be very precise if you have a good model of the material’s work-hardening properties, and fairly fast and low-power as well. But this probably requires annealing the metal (to eliminate unknown work-hardened zones) and possibly keeping it hot during fabrication. Bending of sheet metal is normally done with sheet-metal brakes, press dies, or beading machines. Digital fabrication of press dies is now widespread and can often use cheap plastics.
Drilling metal is pretty high power and imprecise; it may not be very suitable for digital fabrication as a general technique, although digitally controlled drilling might be a useful supplement to other processes, for example to do initial piercing of a workpiece before further processing. Manually drilling metal or wood after holes have been precisely started by a slower digitally-controlled process is also likely useful.
Digitally-controlled grinding should in theory be able to generate very precise geometry (submicron), particularly since the difference between contact and non-contact between a grinding wheel and the workpiece is not at all subtle; it is easy and fast to detect either the light from sparks (maybe 100 microsecond latency) or the sound (maybe 20 microseconds if you have a microphone on the workpiece). This should be able to give a pretty good indication of where the surface of the workpiece is, to within the error of your model of the grinding wheel’s geometry, and you can go back and measure points on the workpiece in between grinding passes by detecting its conductivity. Large grinding wheels are high power when grinding continuously (a US$40 Black & Decker G720N 115mm-diameter 11000rpm angle grinder is 820 W) but not intermittently, and small ones need not be high power or even noisy.
Digitally controlled hammer forging of metal seems potentially very promising, especially since it could produce work-hardened results, but it involves a lot of variables and probably requires annealing steps.
Aluminum foil should be easy to cut to desired shapes under computer control with just a razor blade. If it’s on a softish surface, running a small hard wheel over it under computer control should be sufficient to put ribs into it to give it a little bit of rigidity and define creases for origami panels. Automatically folding the origami or assembling the panels is probably pretty difficult, but doing it manually is certainly practical for initial prototyping. The precise origami shapes thus formed can then serve as tiny molds for inexpensive lightweight castable materials including candle wax, pine pitch, polyethylene cutting boards, and polypropylene bottle caps and rope. If desired, the aluminum foil can then be removed electrolytically with a saltwater or vinegar electrolyte or non-electrolytically with lye or muriatic acid.
Aluminum foil can also be cut with low side forces and tiny holes, but probably lower positional precision, with sparks from a copper point or graphite electrode. Normally you would do this in air but doing it under water or oil may provide better precision.
Mild sheet steel is easy to come by and should be easy to cut precisely but slowly with electrolytic machining with salt water, vinegar, or sulfates. Such cutting can include living hinges that make it easy to bend the resulting cut sheet into precise shapes.
Shapes roughly fabricated in nearly any metal can be brought to precise geometry by electrolytic machining or electric discharge machining, which produce no side forces and no heat-affected zone.
Electrolytic machining of glass may be feasible; see Electrolytic glass machining.
Styrofoam is easy to come by and can be easily cut under digital control with a hot wire. Suitable wire materials that are easy to come by include copper, steel, stainless steel, and nichrome.
Rigid, brittle fine foams such as some of the materials mentioned later offer enormous advantages for digital fabrication: they are as easy to cut to dimension as very soft materials like paraffin wax (though much more abrasive), or even easier, but retain the dimensional stability and refractory properties of the corresponding solid material, which are superior to even some hard metals, and their Poisson ratio is effectively 0; also, and related to their ease of cutting, they are much more resistant to fracture propagation than the corresponding solid materials. The only traditional materials with similar properties are Mykron/Mycalex and exotic mica-containing machinable glass-ceramics.
Corrugated cardboard is very abundant and has a spectacular strength-to-weight ratio and good dimensional stability in two dimensions; it can be cut with razor blades to form precise 2-D shapes and pressed with roller wheels to form precise crease lines, which can then guide origami assembly. Thicker corrugated cardboard may require higher-powered cutting with a wire saw or hacksaw. Cardboard’s worst disadvantage is that it is very vulnerable to water.
Pottery clay bodies are readily available and very easily formed or extruded, although their stickiness can be a problem, and there is currently no software available that can model their behavior well enough to control their forming. If fired, they can produce hard, refractory, dimensionally stable materials with substantial strength, but if not fired, they can easily be recycled.
Digitally guided sandblasting (“abrasive jet machining”) requires compressed air and consumable nozzles, but offers the possibility of cutting desired geometry into a wide variety of materials, even very hard ones, as well as selectively abrading surfaces, exposing fresh unreacted material. This produces such low side forces and heat loads that it can be used even on thin glass or eggshells, though getting it to work on aluminum foil might be too much to ask. Although it even works on metals, it is substantially more effective on brittle materials. Careful choice of abrasives can make this a material-selective process; in industry, for example, glass-bead blasting is used to remove paint without damaging steel, and dry-ice blasting is used for degreasing without damaging paint.
The easiest abrasive to use for sandblasting is of course silica sand, but crystalline silica dust causes silicosis, to the point where sandblasting with silica has been outlawed in many countries. (This doesn’t solve the problem if you’re cutting something like granite, of course, and granite is a very desirable thing to be able to cut because of its dimensional stability.) Sapphire, carborundum, apatite, metal oxides, or powdered waste glass may be preferable abrasives.
A different way to do digitally controlled abrasive cutting might involve passing a fabric loop loaded with loose, perhaps wet and sticky, abrasive across or through the workpiece. This is pretty similar to a wire saw, but I think the flat-tape shape of the fabric potentially provides a better tradeoff between breakage risk and kerf width. To make a given cut, the tension on the leading edge of the fabric needs to be just as high as it would be without all the cloth behind it, but if that turns out to be too high, the rest of the cloth doesn’t have to rip as well.
Digitally guided 2-D plasma or oxy-fuel cutting is of course already a widespread process, and is much faster than most alternatives; laser cutting is making significant inroads here in recent years, but would be much more difficult to improvise.
The “Oogoo” Sugru-like silicone putty mix of cornstarch and hardware-store silicone caulk is thixotropic with a working time of minutes to hours (5 minutes with a 1:2 ratio, an hour with a 5:1 ratio), and thus eminently suitable for digital fabrication through extrusion or forming.
Plaster of paris can be foamed with baking powder, then cut to shape. It adheres well to quartz sand, and quartz sand or quartz flour can be useful functional fillers for it.
Charcoal is an easily cut material related to carbon foam (see below), but it tends to suffer pervasive cracking from thermal contraction during its formation, which may limit its uses.
A lot of the processes in the previous section for converting digital data into a physical three-dimensional shape are only applicable to a narrow range of materials, often with fairly poor properties for anything besides being shaped. So it’s important to be able to transfer a geometry fabricated in one material into some other material.
Hardware-store silicone caulk is a promising material for molding for a few different reasons. It’s capable of holding detail down to the micron level; it’s fairly inert once set, withstanding, for example, gasoline; it’s thermally stable up to typically 300° (though the red high-temperature formulation commonly used for auto repair as “RTV” goes a bit higher); it’s fairly dimensionally stable (though it does shrink a little, unlike some other silicones); and it’s elastomeric. Being elastomeric makes it easy to pry it out of molds made of a harder material, and also makes it easy to peel molds made out of it off castings of a harder material. Also, it won’t remain stuck to polyethylene or polypropylene, and reportedly also not to PVC or polycarbonate.
Shapes initially fabricated in aluminum foil can be thickened by electroplating/electroforming them, most easily in copper or brass. Although it is important for the copper to form a solid layer, it is not necessary for it to adhere firmly to the aluminum, as is usually desired in electroplating processes. Because the aluminum foil is the cathode in the electrolytic cell, it is possible to use electrolytes such as muriatic acid or lye that would normally destroy it immediately. A potentially larger concern is the risk of deformation from the surface tension of water, which can be reduced with surfactants and the substitution of alcohol for water.
Electroforming on non-conductive shapes is conventionally done with graphite powder dispersed in some solvent and painted onto the object. In some cases you can disperse the graphite in a solvent that softens or dissolves the surface of the object, welding the graphite to it; dichloromethane is reported to work for PLA.
Polypropylene can be sufficiently stiff for molding of plaster of Paris or portland-cement concrete, which can provide a polypropylene part with much greater rigidity and thermal stability than polypropylene alone. In the case of plaster, it may be possible to later strengthen the plaster by filling internal channels in it with molten metals such as aluminum, brass, or cast iron.
Styrofoam forms, even those that can be cut by hot wires, can be stacked up into forms for low-temperature molding (for example, of portland cement or plaster of paris), or they can be wrapped in papier-mache-family strengthening materials such as cotton cloth soaked in plaster of Paris or fiberglass window screen soaked in non-alkaline sodium silicate.
Plaster shapes have little tensile strength or rigidity, but if they are suitably designed, automatically winding them in pretensioned wire can improve this. Steel wire has higher rigidity and less creep than copper wire but may be harder to find; aluminum wire is available by dissecting window screens, or the woven screening can be applied directly. Winding subsequent layers at different angles, as is done for glass-fiber pressure tanks, can provide tensile strength in two dimensions instead of just one. Brazing or soluble silicates may be suitable means for obtaining adhesion between layers of winding.
XXX stuccoing
A lot of attractive fabrication processes, such as firing clay, require high temperatures at some stage; so, too, does making many exotic materials (see section below). Making equipment that survives these temperatures requires refractory materials, often insulating refractories, although in some cases it’s adequate to just use a pile of quartz sand (good up to 1500°, though not very insulating) or vermiculite (insulates better, I think good to 1100°). Aluminum foil can’t resist high temperatures itself but is often useful for reflecting back radiant heat, preventing it from being lost.
Ordinary steel works up to about 1200° in a reducing atmosphere, but carbon dioxide is not sufficiently reducing; in air it starts to oxidize annoyingly rapidly above about 900°. Fired clay is the usual resort for temperatures up to 1100° or so; special clays can reach much higher than this but are harder to find and to fire. Fused quartz is maybe sometimes good to 1500° and usually to 950°, and is available for example in broken space heaters and halogen light bulbs, but it’s very difficult to cut or form. (In some cases the quartz tubes are adequate.) Plaster of paris is easily formed before hydration, and can withstand a few excursions to over 1000°, but is not durable as a refractory.
Soluble silicates are hard to find (see below about making them), but can serve as adhesives for silicates such as quartz. Typically, in this use, rather than melting at high temperatures and falling apart, they form new compounds with the materials they’re uniting.
Carbon foam is an excellent insulating refractory in non-oxidizing atmospheres (good to 3642°) and can be fabricated easily from bread dough or pancake batter, which is first heated to dry it and make it rigid, then heated further to carbonize it. It is very rigid and thus easy to cut, but abrasive. It does not adhere well to untreated quartz fillers. Thermoplastics alone are not suitable precursors for carbon foam; enough thermoset ingredients such as gluten are required to prevent the object from losing its shape before carbonizing. If heated sufficiently in a non-oxidizing atmosphere it may graphitize and become electrically conductive, depending on its structure. Carbon dioxide is sufficiently non-oxidizing.
The standard insulating refractory for low-tech pottery kilns in an oxidizing atmosphere is a conventional pottery clay body (for example, ball clay tempered with silica and grog) filled with particles of a sacrificial-filler organic matter such as coffee grounds, sawdust, or used yerba mate, which burns out upon firing. In my experiments, material made with 67% sacrificial filler was quite solid but could be cut with a thumbnail, while material made with 89% sacrificial filler was still solid but friable and permitted easy gas passage. The firing process produces terrible odors.
Intumescent moldable “Starlite”-style coatings may be adequate insulating refractories for bootstrapping high-temperature capabilities. The precursor is an aqueous paint or paste of organic polymers (such as cornstarch and PVA glue, or wheat flour) and blowing agents. Sodium bicarbonate is commonly used as a blowing agent. Borax or boric acid substantially increases the strength of the resulting carbon foam, and may also help to cross-link PVA in the paste to prevent cracking from drying.
Silicone caulk may work as a precursor material for composites of graphite and carborundum, foamed or not, when heated. Acetic-cure silicone may cure more rapidly and foam in the process if carbonate or bicarbonate of soda is mixed in; I have not verified this. Oogoo confirms that it does cure more rapidly when mixed with cornstarch.
Even if you have an apparatus that can withstand heat, where do you get the heat?
The traditional approach for millions of years has been fire. Ordinary butane blowtorches can hypothetically reach 1970°, but usually don’t, which is why you can’t weld steel with them, and they have the inconvenience of producing a lot of exhaust. Oxy-acetylene torches are easy to buy (though expensive to refuel) and can reach 3500°, and oxyhydrogen torches are easy to make and can reach 2800°. Anthracite, and thus presumably charcoal, can reach 2180°.
But electric heating is much more convenient; it can be turned on or off (or anywhere in between) instantly, and it doesn’t produce gases. Ordinary nichrome heating elements have maximum service temperatures ranging from 1000° to 1260°, though they don’t melt until almost 1400°. Some varieties of Kanthal have service temperatures ranging from 1300° to 1425°, but these are harder to find. Halogen lamps, still available from auto parts stores as headlights even where they’ve been prohibited for household lighting, and their filaments may reach 2900°, but their envelopes are only designed to operate around 500° and are typically made of fused quartz, which melts at 1600°, or aluminosilicate glasses, which melt at only about 800°.
Carborundum “globar” heating elements are commonly rated to 1625° or 1600°, but are also not common household items, though it might be possible to make one; they consist of a carborundum tube with a spiral cut in the central portion to increase its resistance, so that the ends that protrude through the refractory wall of the furnace can remain cool enough not to melt the metal wires that connect to them. The “Globar SR” design has a two-start spiral cut so that both electrical terminals are on the same end. Carborundum is seriously allergic to water vapor.
Historically, the fairly expensive yttria-stabilized zirconia was also used for globars; they melt at 2715° and have been experimentally used for heating up to 2100°. Possibly household ceramic knives could be used for this, though they might need to be cut to have a central “hot zone”, similar to carborundum globars. One disadvantage is that they need to be preheated (for example with a flame) to become conductive; historically, their negative temperature coefficient of resistance was also a drawback (for example, in Nernst illumination lamps), since it means they require a constant-current source rather than a constant-voltage source to avoid thermal runaway. (Carborundum, by contrast, has a positive TCR above 700°, so this issue doesn’t arise at normal globar service temperatures.) Nowadays current regulation, at least, is an easy problem to solve.
Nowadays, high-temperature heating elements are commonly instead the exotic MoSi2 instead, which is serviceable to 1750° to 1850°. These are commonly used, for example, for sintering zirconia itself, which commonly requires 1530°-1700° depending on, among other things, sintering aids.
Alternative methods of electrical heating include arc heating with consumable graphite electrodes and induction heating, neither of which has an inherent temperature limit of its own; arcs in everyday US$200 plasma cutting torches commonly reach 20000°. Induction heating can keep the induction coils outside the hot furnace, and big induction heating coils are commonly made from copper pipe (at the high frequencies used for metals above their Curie point, only the skin of the coil can carry current anyway) with cooling water running through it. Induction furnaces in industry commonly maintain metal molten by heating the liquid metal inductively.
XXX microwave heating
Teflon and glass are crucial materials for their nonreactivity at everyday temperatures. Glass is widely available and, though it requires a lot of practice, can be shaped with a US$10 butane torch from the hardware store (or, traditionally, with an oil lamp and a blowpipe); teflon can be obtained from discarded laser printer fuser rollers, and a great deal of electical insulation is also made of teflon, but I do not know how to distinguish teflon insulation in discarded cables from the more common PVC.
Graphite is a crucial material for both electrodes and crucibles, the only viable electrode or refractory for many purposes; welding shops sell graphite electrodes, but they are graphite composites with poor stability in reactive environments. As mentioned above, some organics can be graphitized in a graphitizing furnace made of carbon foam and purged with carbon dioxide. This requires, I think, electric heating elements that can withstand graphitizing temperatures of 3000° (and pure graphite itself is the only plausible option), but are more conductive than the carbon foam itself. Even if the carbon foam is made from non-graphitizing carbon, it will conduct electricity once fired high enough.
Non-graphitizing carbon crucibles, which are more resistant to reactive environments than graphite, have been historically made from phenolic resin, then fired at 900° in an inert atmosphere. Other thermosets would presumably work too, unless they pyrolyze to graphitizing carbon (polyurethane foam, as found in pillows and spray-foam insulation, is a common precursor); if they don’t outgas too much they might be able to make non-porous non-graphitizing carbon.
Soluble silicates, especially those that are neutral rather than strongly alkaline, are likely a crucial enabling material for digital fabrication for several reasons: they can be used directly as refractory adhesives (for example, to make a moldable insulating refractory from garden-store vermiculite); they bind very strongly to silica and other silicates; when dehydrated to solidity, they can be expanded from beads into glass foams by the application of heat, foaming as their water boils; and they can be instantly and directly cross-linked into insoluble silicates by the provision of polyvalent cations such as calcium or ferrous ions, as in the traditional colorful silicate garden, or with carbon dioxide, a feature that promises to be important for digital fabrication by selective solidification, perhaps even permitting 3-D printing of soda-lime glass. But soluble silicates are difficult to find, so we may have to make them.
(It may be possible to leach neutral sodium silicate out of corrugated cardboard.)
The most promising route to soluble silicates seems to be the digestion of powdered soda-lime glass with warm aqueous alkali over the course of hours or days. Alkali requires quite careful handling and can itself be difficult to obtain; the chlor-alkali process with graphite electrodes and a porous fired-clay diaphragm can produce it from table salt, salt-substitute potassium chloride, or alkali carbonates, and producing it thus in situ may be adequate for digesting the glass, thus avoiding any accumulation of hazardous alkali. Unwanted chlorine may be disposed of by passing it over hot aluminum foil. The traditional source for alkali is to leach it out of wood ash, but this is normally slow, dirty, bulky, and expensive.
Although discarded soda-lime glass is abundant, it doesn’t handle thermal shock or reactive environments well; borosilicate glass is much more resistant, but hard to find. Adding borate (in the form of borax or boric acid) to discarded soda-lime glass seems like a promising thing to try.
Sapphire is aluminum oxide; industrially this is produced in the Bayer process by digesting bauxite with alkali to produce a soluble aluminate, then precipitating gelatinous aluminum hydroxide by cooling the solution (neutralizing it also works). This hydroxide (which crystallizes as gibbsite at a few microns per hour) calcines to sapphire, completely if held above 1200° for an hour; it is even possible to calcine the hydroxide gel to a transparent porous ceramic at 500° if you keep the electrolyte concentration near an ideal value, though perhaps not if the hydroxide is derived in this way. Digesting (readily available) aluminum metal with alkali produces the same soluble aluminates, and so should be an easy route to sapphire for use as an abrasive, as a refractory (melts at 2072°), or for abrasion-resistant ceramics; the sapphire powder sinters to a ceramic around 1600°. Its thermal coefficient of expansion is an astounding 0.6 ppm/K at ordinary temperatures, though some sources give higher value such as 7 ppm/K.
The so-called “sodium beta alumina” that forms when heating sodium-rich gelatinous (?) aluminum hydroxide has fast ionic conductivity for a wide variety of monovalent cations, a property of great interest for its use as a solid electrolyte (and one which may be much more accessible than zirconia), used in the sodium-sulfur battery.
The transformation sequence from the gelatinous hydroxide to sapphire (alpha-alumina) is astoundingly complex (8, figure 4.1); the gelatinous form converts to the poorly ordered eta-alumina (aka gamma-alumina or gamma-prime alumina) around 375° (or 626°?), which converts to theta-alumina (“a better ordered transition form”) around 800°, which finally converts to sapphire around 1120°, while if instead it is first crystallized as gibbsite (as is usual in the Bayer process) some of it instead goes by way of chi-alumina and kappa-alumina. Six other oxide and hydroxide forms are also potentially part of the process, depending on impurities and heating rates. Contamination with carbon dioxide in this process may result in incorporation of carbon.
The gibbsite itself is an important functional filler for plastics, providing strength and especially fireproofing.
Lucalox is another potentially important use for sapphire.
Sapphire can also be crystallized hydrothermally with soda ash above 400° and 200 MPa, but such pressures are challenging.
Another interesting material potentially derived from aluminum hydroxide is mullite, the acicular aluminum silicate that accounts for the legendary refractory performance of Hessian crucibles. Crystallization of mullite from amorphous Al6Si2O13 at 980° has been reported; the preparation was a difficult process involving alkoxides of aluminum and silicon (and the phrase “was put in an open flask and stirred for three months”), but since mullite is the stablest alumina/silica compound, perhaps easier routes exist, such as just firing at 1100°.
Carborundum was initially discovered by heating sand in an iron crucible with an arc from a carbon electrode submerged in the sand, carbothermally reducing some of the silica with some of the carbon from the electrode. Heating mixed carbon and sand with a submerged arc sounds easy, and you don’t even need the iron crucible; you just need two carbon electrodes.
HNO3 was traditionally obtained by XXX
Above I presupposed we could get precise positional control. But how could we get such positional control? Norbert Heinz has demonstrated a series of excellent homemade CNC machining tools made from hardware-store parts, using the gantry arrangement used by most existing CNC machines. Some of them use H-bridge-controlled DC motors (or steppers) and optical quadrature encoders he has cut out of tin cans (or from paper) with optical sensors from old printers. The shaft rotation is translated to linear motion with leadscrews from hardware-store allthread; but this feedback measures only the rotational position of the leadscrew, so it is subject to errors from deformation of the machine frame and the leadscrew and from backlash, as well as thermal expansion.
To avoid these errors, you need feedback about the actual position of the end effector rather than motors that indirectly drive it. In theory optical mice have a resolution of a micron or two, but Heinz found they lose steps. He has achieved more precise control using the transparent plastic optical encoder strip from an inkjet printer to measure the linear position; these are usually 92, 150, or 300 lines per inch. I can’t find Heinz’s page on the topic, but in 02010 Michele Lizzit reported 33-micron precision using this method.
Industrial machine tools are now almost universally equipped with a “DRO”, digital readout, or are fully automatically controlled. Three common kinds of DROs exist: using optical “glass scales” similar to the inkjet-printer encoder strip, using magnetic sensors that read a strip of alternating magnetizations, and using capacitive sensors that read a strip of alternating electrical connections. Digital calipers with 100-micron precision using this capacitive system are widely available at retail for about US$8, and are commonly equipped with an easily-tapped internal SPI data bus, while the other two systems routinely deliver micron precision.
These approaches can suffer from thermal error when the scale (glass, plastic, or otherwise) expands or contracts under the influence of temperature variations. The traditional solution to this was to keep the temperature of your metrology lab constant to within a tenth of a degree, but an alternative is to use laser interferometry, which can easily deliver submicron precision and is much less affected by temperature.
The kind of swing-arm arrangement used by manual magnetic 2-D profile cutting machines, or by hard-disk platter-and-arm arrangements, is mechanically vastly superior to the gantry arrangement; Melisa Orta Martínez’s “Haplink” design demonstrates a promising mechanical design for adapting such a swing-arm arrangement to motor-driven cable drives.
For precise actuation over short distances, flexures and voice-coil actuators are probably the best approach. Hard disks have voice-coil actuators in them.
Differential roller screws ought to enable far more precise linear actuation than currently popular systems, but without digital fabrication, they are very expensive to build.