In Dercuano I wrote a bit about a family of likely-feasible powder-bed 3-D printing processes where you print the shape by selectively depositing some “flux” in a powder bed (for example, by inkjet deposition of a dissolved aqueous binder, or by trickling granulated binder from a vibrating chute or screw extruder, like making a sand painting), then bake the whole powder bed to activate the flux, forming a solid object. I had done a few manual tests with 100-micron quartz flour and various different candidate fluxes fired at IIRC 1040°, getting some promising results.
Well, now I have a lower-temperature version of the process that I have a lot of confidence in, although I haven’t tried it.
The powder bed consists of sieved pulverized soda-lime glass, which is one of the candidate fluxes I’d tried at the higher temperatures. The flux is lye. The lye becomes very reactive when heated past its melting point of about 320°, capable of attacking soda-lime glass at corrosion rates millimeters per minute, and forms waterglass with the surfaces of the glass particles, which is less reactive and has a higher melting range than the lye; it bonds the glass particles together into a solid mass. This is too viscous and has too much surface tension to infiltrate the rest of the mass, so unfluxed glass particles remain inert; soda-lime glass generally doesn’t soften below some 700°, so at this baking temperature, the glass won’t stick together.
Generally molten lye is considered a material meriting the sort of respect we accord to molten steel, lava, or RFNA (unless we’re cleaning our ovens, in which case we often treat it casually), but in this case we’re dealing with very small quantities which exist for only a short time within the hot powder bed. If the glass granules have a size on the order of 100 microns, they might each occupy 0.6 nanoliters and be associated with an additional 0.4 nanoliter void space. If the lye is initially deposited as a 1-molar solution, that’s about 4% by weight or 3% by volume, so we have about 0.01 nl of lye for each 0.6-nl glass granule, distributed among the “necks” connecting it to its neighbors; even the mass in the fluxed zone is 98% inert glass and 2% molten lye, and the overall powder bed might be more like 0.2% molten lye and 99.8% inert glass. A 1-kg printed glass piece might include some 20 g of lye before firing.
After baking the powder bed, first at 100° long enough to drive off any water and then at 350°-500° to sinter the lye-fluxed glass particles, depowdering the finished object should be straightforward, and the leftover powder can be sieved and reused. The total kiln time should be on the order of an hour, more for thicker powder beds through which the heat will propagate more slowly; slow heating and cooling may also be necessary to prevent breakage. Pressure will not be needed to cause sintering, and may not be effective, since the rigid support from the unfluxed particles will prevent any significant compaction of the fluxed particles.
Although I’m now pretty confident that the above will work, there are a number of variations that might be better; some are sure things, others less so.
If the lye is deposited as a liquid solution, inkjet-style, it may be helpful to use ethanol as part or all of the solvent to reduce the surface tension and thus the size of the binder droplets, as well as accelerating evaporation. To prevent it from spreading out through the powder bed once deposited, it may be helpful to either use a very high lye concentration, like 50% (12M), or to include alkali-tolerant thickeners (especially thixotropic thickeners) to get high viscosity and avoid filling the void spaces in the powder bed. Such thickeners would become part of the final workpiece, which probably rules out most organic chemicals, since they would char and turn the glass black.
Being able to use lower lye concentrations in the flux would be valuable because concentrated lye solutions are super annoying.
Of course the most obvious thixotropic alkali-tolerant thickener to use would be waterglass itself, in which case you don’t need the lye at all and may not need the baking step either. Exposure to carbon dioxide, either as a gassing step after printing or just from the atmosphere, may provide adequate strength.
Fused silica, silica gel, or silica fume (perhaps granulated) may be a more suitable powder base than the cheap-as-dirt soda-lime glass discussed above. Their absence of alkaline-earth metals enables them to react more readily with alkalis, they have much lower thermal coefficients of expansion, and they do not sinter on their own until higher temperatures, perhaps 1000°-1600°. But the resulting soda-silica glass is much more vulnerable to corrosion from water than conventional soda-lime glass. You might be able to add borax to the flux, perhaps reducing or eliminating the lye, to get a borosilicate glass instead of a soda-lime glass. But you need a lot of boria for borosilicate glass, like 8%-15%, and borax only dissolves at like 25 g/l at room temperature, up to 250 g/l at 70°. Using very porous powder bases like silica fume or dehydrated silica gel might ease this constraint, but will tend to produce a lot of porosity in the final print.
Incorporating desiccants into the powder base is another possible way to prevent binder droplets from spreading out once deposited; they don’t need to be stronger desiccants than the lye, just strong enough to diminish the free liquid volume somewhat and keep the lye from escaping. Promising desiccants for this purpose include silica gel itself, activated alumina (incompletely calcined aluminum hydroxide), quicklime, and zeolites.
A totally different powder base that perhaps could form waterglass with lye more easily than soda-lime glass or quartz sand (see below about the crude sand experiment) is cristobalite; Hachgenei et al.’s US patent 5,215,732 explains that at room temperature, cristobalite (or “tempered quartz sand” containing a mixture including cristobalite, the other high-temperature polymorph tridymite, and amorphous silica) is so much more reactive than quartz that you can convert it completely into waterglass by boiling it in 50% lye at 112°-146° for only three hours, even with a 2:1 ratio of silicon to sodium (“modulus”). Higher ratios of sodium to silicon naturally run faster. He gives the sand grain size as “in general, ... 0.1 to 0.8 mm,” so the corrosion was advancing toward the center of the sand grains at around 30 nm/s, times or divided by four. He reports being able to thus “temper” quartz by calcining it “above 1000° C., preferably at 1300° to 1600° C., with the addition of catalytic quantities of alkali”, and, presumably, not annealing it back to alpha-quartz.
My original flux-deposition notes contemplated depositing the flux as a powder, for a fully dry 3-D printing process. It’s desirable to use as fine a powder as possible, because the grain of the powders limits the dimensional precision of the workpiece; even if you deposit your flux with 10-micron precision, each 100-micron-wide flux particle that becomes part of a surface is going to produce roughness on the order of 50 microns. Similar concerns apply to the powder-bed particles: if your flux particles are sticking together 100-micron-wide particles of aggregate, you’re going to get surface roughnesses on the order of 100 microns. But using fine floury powders for the powder bed isn’t problematic; however, fine floury powders of flux tend to clump up into larger lumps, as the surface forces between particles overwhelm their weight and inertia.
I’ve written about various approaches to solving this problem in Precisely measuring out particulates with a trickler, but a new one occurred to me in this context. If you can blow the flux powder through a nozzle with a compressed air blast, you can do an excellent job of breaking up aggregates, because the aerodynamic forces on the particles tend to scale with their surface area, just like the surface adhesion forces they’re fighting, not with their volume. So you can blow very fine dust through a quite fine nozzle, although once you’re down in the sub-ten-micron range you start having safety concerns. But of course if you’re blowing that air blast onto a powder bed, those same aerodynamic forces will tend to disrupt the powder bed, which is why I’d never considered this solution before.
The objective here is for the powder bed to be an unbound, loose powder after baking, except in the places where enough flux was deposited to enable it to bake to a solid. So you’d think that including a binder in the powder bed would be totally counterproductive. But if we use a sacrificial temporary binder that bakes off during the baking process without leaving a residue, it could still work!
Suitable sacrificial temporary binders might include water (as used in
traditional fired-clay ceramics), ethanol, gasoline, kerosene,
toluene, turpentine, d-limonene, acetone, ethyl acetate, dimethyl
sulfoxide, isopropanol, formamide, hydroxylamine, sulfur, naphthalene,
and various salts of ammonium (chloride, carbonate or bicarbonate or
carbamate, acetate, sulfate); most of my notes on these are in file
inorganic-burnout.md in Derctuo.
Some of these candidate sacrificial binders, like water and ethyl acetate, are ordinarily liquids, which means that the powder bed is sort of more of a paste bed; the powder base might be premixed with the liquid, so that the recoater trowels on one layer after another like drywall mud or construction mortar. This could lead to geometric disturbances from subsequent recoating layers, though that might be solvable; the usual practice with construction mortar is to minimize this problem by including barely enough water in your mortar to make it plastic, and as soon as you trowel it onto the wall, it loses enough water to lose its plasticity.
Alternatively, the recoating could be done with a powder recoater in the usual way, which is gentle enough not to perturb previous layers, then misted with the sacrificial binder liquid.
Other candidates, like ammonium chloride and ammonium carbonate, are ordinarily solid, so if the base powder particles are bonded together by them, it becomes a solid object, which avoids the issue of geometric disturbances when recoating, but poses the problem of how to get the sacrificial binder into the powder. If it’s just mixed in as separate particles, it won’t be form bonds between the base powder particles that are adjacent in their final position. Ammonium carbonate is water-soluble; it could be sprayed onto the base powder once it is in position, or the base powder could be mixed with an aqueous solution of it, and in either case we would need to evaporate the solvent to get to solidity. Ammonium chloride has an additional power: it can be “sublimed” at convenient temperatures, so it could be analogously infused into the powder once it is in place without needing a liquid solvent. Or it can be formed in situ by reacting ammonia with muriatic acid.
See Powder-bed 3-D printing with a sacrificial binder for more variations on the theme.
Some candidate temporary binders, like muriate of ammonia, can be produced in situ in the powder bed by applying two different reagents at different times; in that case, ammonia and muriatic acid can be infiltrated into the powder bed in gas form, one after the other, where they will react to form the salt.
Some of the alternative powder-bed bases such as cristobalite or amorphous quartz offer the possibility of using ordinary muriate of soda as a less annoying alternative to lye that activates at higher temperatures. It has been used for salt-glazing of pottery for 600 years. At 1100°-1200° in a steam atmosphere it forms muriatic acid gas and lye; the former escapes, driving the reaction forward, while the latter fluxes the silica as before.
Wikipedia claims that silicates of iron are even more effective fluxes for this purpose, so reduced (ferrous) iron helps even more, and red clays are well-known to be lower-firing; however, ferrous silicate on its own is fayalite olivine, and olivine is the canonical high-melting mineral at the high-temperature extreme of Bowen’s reaction series. But I think fayalite olivine may indeed be low-melting; a 01993 abstract in Science gives its melting point as 1478 K, which is only 1205°. So some ferrous iron in the powder bed may help the process along.
Higher-temperature powder-bed bases might also be able to use soda ash as a flux. Soda ash is water-soluble, even cheaper than lye, less annoying to handle, and melts at only 851°. My experience melting it with a butane torch suggests that it has an alarming tendency to bubble, perhaps because it is slowly converting into lye. This produces only carbon dioxide gas rather than muriatic acid.
If iron silicates are crucial to the fluxing effect in the salt-glazing of pottery, perhaps metallic iron or some iron salt could form these iron silicates directly with soda-lime glass at lower temperatures than those necessary for salt-glazing. Ferric chloride melts at only 308° but has a very narrow liquid range, while green vitriol starts to decompose into high-melting hematite at 680°.
If transparency of the glass produced is not a concern, wood ash might be a possibility; it would surely be the cheapest flux for powdered glass, if it works. It is mostly carbonates, oxides, and hydroxides of alkali and alkaline earth metals, including lye; leaching out the water-soluble components will tend to eliminate the counterproductive polyvalent cations. Traditionally this was done by washing the ash on top of linen cloth, then boiling down the results to reasonably pure potassium hydroxide.
I did a crude and deadly kitchen experiment the other day which seems to have successfully made a little waterglass from quartz and lye. I placed a layer of damp construction sand in a thin stainless (or nickel-plated?) metal bowl, sprinkled a layer of lye flakes liberally on top of the center of the layer of sand, then pressed down another layer of damp construction sand on top. I covered the bowl with aluminum foil, placed a paper towel over the aluminum foil, added an aluminum-foil skirt around the edge to reduce the loss of radiant heat from the bowl’s sides, gently heated it on a gas stove burner (maybe 500 W) until the lye flakes stopped crackling from the release of water. Then I turned the burner up to max (maybe 1500 W) for an hour; halfway through I moved the bowl over a bit to ensure that there were no cold spots that never got heated. I left the sliding-glass door open so that the draft would carry any fumes from the bowl away from me.
Unfortunately I don’t have a thermometer. Some of the aluminum foil skirt around the sides of the bowl strayed into the gas flame and melted, but on the various occasions during the hour when I inspected the crude apparatus, no part of the bottom of the steel bowl itself was ever glowing visibly, so it had not reached the 525° Draper point. When I turned the flame off, tore open the aluminum foil, and poked the sand with a chopstick, the end of the chopstick charred and smoked, but didn’t burst into flames, so the top of the sand was probably somewhere between 250°-350° at that point. The sand had formed a hard mass, infiltrated by the molten lye, although not, as it turned out, around the edges of the bowl; only in the middle. Presumably the sand in the bowl was hottest on the bottom where it was separated from the flame only by a thin layer of steel, with a net heat flow in at the bottom and out at the top producing a thermal gradient through the sand.
Upon cooling I was able to use the chopsticks to lift up a large monolithic aggregate of sand that extended all the way to the bottom of the bowl; at its bottom and top surfaces its color was the beige of the sand, but in between, where the lye flakes had been, it was much more white. No intact lye flakes were in evidence; they had all completely melted, so that part of the sand had exceeded 320°.
I broke off a small piece of the aggregated sand with the chopsticks and dropped it into a polypropylene bottlecap to which I added a few drops of water; it disintegrated immediately, indicating that the binder was probably mostly lye, not mostly waterglass. A couple of flakes of aluminum foil added to the bottlecap fizzed enthusiastically, confirming the presence of free lye. So at this point I had no indication that any waterglass had been formed.
I neutralized the rest of the sand by soaking it with kitchen vinegar; after letting it stand a while, I added a pinch of baking soda, which fizzed, confirming that the pH had been brought down below neutral. This ensured that any lye had been not only dissolved in water but converted to highly soluble sodium acetate; at the same time, the lower pH would make any waterglass present almost entirely insoluble in water.
Stirring around the sand with my fingers, I found that most of the aggregated chunks had disintegrated immediately upon wetting or were easily broken up. However, one irregular chunk of aggregated sand of a centimeter or three in every dimension remained intact, indicating that it was completely bound together with something other than frozen lye, almost certainly waterglass. Handling it wet did not leave my fingers slippery, providing further confirmation that free lye was not present.