We’ve been doing some experiments in 3-D printing of glass parts. It’s a classic powder-bed binder-deposition process, except that the binder is another solid powder, and the postprocessing is different — rather than removing a fragile greenware part from the powder bed and then possibly baking, curing, or soaking it, we fire the whole powder bed to activate the binder, which acts as a flux to sinter or even dissolve the surfaces of the powder grains; the binder itself may or may not melt.
We’ve had success so far with a quartz powder bed and fluxes of soda-lime glass frit and a commercial pottery glaze, both baked at 1020°, in air, at atmospheric pressure. We also expect success with fluxes of potassic feldspar and lead oxide. In all of these cases, the final product is either a glass or a glass with a quartz filler.
However, this process is much more widely applicable.
The flux-deposition process, unlike most binder-deposition powder-bed processes, involves no extra step of handling delicate greenware extracted from the powder bed. This extra step is labor-intensive and often involves extra equipment such as sandblasting cabinets to remove extra unused powder without damaging the greenware part.
Aside from the reduction in human effort, this should improve the resolution. Typical binder-deposition processes as offered by service bureaus like Shapeways cannot reliably produce walls and wires under about 1000 microns in thickness. By contrast, flux deposition powder bed printing should offer very high resolution — 3D Systems’s direct-metal selective laser melting machines can, as of 2015, produce parts with 100-micron wall thickness because they, too, have no greenware-handling step, so the articles are already at their full strength when they are removed from the powder. We think flux-deposition printing can offer the same or better resolution, but without the long build times and inert atmospheres needed for direct metal SLM or SLS.
While we do not expect flux deposition to be able to avoid the issue of dimensional change during sintering, which is a plague for sintering manufacturing processes in general, there is good reason to expect the distortion introduced by the process to be very small compared to fused deposition modeling, selective laser melting, selective laser sintering, or machining of green ceramics before firing them. The binder is deposited when the powder bed is at room temperature, so the average expected final expansion is zero; the article is supported by unused binder throughout the process, limiting forces that could produce deformations; and the process need not involve any strong thermal gradients.
Immediately after sintering, the boundaries between the original powder grains will be enriched in the flux, while their cores will contain none of it; but if the flux and the bulk material are miscible in solid solution as well as liquid, then diffusion annealing can remove these inhomogeneities. This process can allow the article thus produced to withstand higher temperatures than those used to make it in the first place.
Aside from the microscale inhomogeneities stemming from the union of different materials in powder form, it’s possible to use this process to create macroscale inhomogeneities; for example, by smoothly varying the concentration of lead oxide in a glass article, it should be possible to fabricate an article whose refractive index is not constant, but has some controllable gradient. If the microscale inhomogeneities can be eliminated, such graded-index optics can constitute monolithic refractive optical systems that approach diffraction-limited performance arbitrarily closely, lacking as they will refractive index discontinuities to reflect stray light.
Iron melts at 1538°; due to alloying with 0.18% of carbon, AISI 1018 steel, the most common steel, melts around 1520°. AISI 1070 steel is popular for harder goods such as knives; it contains 0.70% carbon and melts at 1479°. The iron-carbon phase diagram shows a eutectic point at 4.3% carbon at about 1175°, and some cast irons approach that content of carbon, typically replacing some with silicon.
This process is already used industrially to produce sintered steels; for example, Höganäs offers a range of “pure iron powders” for this purpose, which all range from 20–200 μm in particle size, with little variation in that dimension.
As of 1997, iron-powder metallurgy normally takes place at 1120°–1150° for 20'–30' because this allow mesh-belt furnaces to survive, although some processes supposedly may require a sintering temperature between 1100° and 1300°. Even 1120° is enough to sinter plain iron. Carbon is added as graphite; endogas avoids removing the carbon, though we think mere carbon dioxide should be adequate.
Apparently it is common to include up to 2.5% powdered copper in the iron in order to fill pores and produce precipitation-hardenable alloys. So copper is a possible alternative to carbon for this purpose; copper and carbon are usually used together to avoid thermal expansion induced by the copper.
The biggest difference with the process being considered here is that iron powder is normally compacted into a mold at 200 to 800 megapascals before being sintered.
Iron phosphide, though hazardous to handle, is another alternative flux which is known to form a eutectic with iron at 10% phosphorus that melts at only 1050°. The resulting iron-phosphorus alloy is similar to a steel, with final phosphorus concentrations of up to 0.6% being beneficial. (Phosphorus is considered a contaminant in normal steelmaking for reasons that are not relevant here.)
Aluminum is probably not a good bet; it forms no eutectic with iron, its miscibility is low, and the intermetallics are brittle and weak, although there has been promising work by Hansoo Kim in Korea to fix this with manganese, nickel, and carbon.
This makes us optimistic that even a very small amount of carbon, copper, or iron phosphide could work effectively to flux a powder bed of pure iron heated to somewhere below 1120° so as not to sinter the iron itself. If the carbon particles are small enough relative to the firing time, they will dissolve entirely into the iron particles; further diffusion while hot should gradually eliminate the lowest-melting-point liquid regions. Depending on the size of the iron particles, the result should be either a pure steel article, or an article composed of grains of pure iron cemented together with steel or cast iron, in a bed of loose iron powder.
We don’t know if there is a way to get the articles to be fully dense; this would require that the steel or cast iron have greater volume than the iron and carbon that went into them. This seems unlikely, as steels are usually slightly denser than iron.
Korol’kov & Kibak report that iron sinters more easily with 100–250 ppm boron; they mention that the eutectic liquid phase Fe₂B + Fe melts at 1160° to 1180°, and they sintered at 1160°. Also, their 100ppm boron-doped pressed iron powder began to shrink at 450°–500°, suggesting that sintering was already beginning, and hit a discontinuity at about 900°. They suggest that this initial shrinkage was due to cementation of iron grains with liquid oxides of boron.
Aluminum natively melts at 660°, but for casting, it is normally alloyed with silicon, with which forms a eutectic at 12.2% Si, melting at 577°. Silicon itself melts at 1414°. Aluminum-alloy powders are commonly sintered at 590–620°, which is hotter than the melting point of the eutectic. It seems that it should thus be possible to flux an aluminum powder bed with small amounts of silicon, bake the result somewhere between 500° and 590°, and get good “cast” aluminum articles out without sintering the aluminum powder.
Sapphire, Al₂O₃, also known as corundum, aluminum oxide, or ruby, doesn’t melt until 2044° or so. It’s an extremely hard and refractory material widely used for engineering ceramics as well as bulk refractories.
In 1957, Cutler, Bradshaw, Christensen, and Hyatt showed that the addition of about 4% (wt%) of appropriate fluxes could lower the sintering temperature of finely divided sapphire to 1300° to 1400°. Specifically, they added oxides of manganese, titanium, and copper as fluxes. 2% of either MnO or Cu₂O and 2% of TiO₂ were successful at getting very nearly full sintering to 3.8 g/cc at 1300°. Furthermore, they reported “bleeding” of a black liquid from their pressed pellet specimens onto the coarse sand below, suggesting that the mixture didn’t merely sinter — it partly melted. This suggests that sintering should be possible at even lower temperatures, although they did not attempt this.
Indeed, Xue and Chen successfully sintered alumina at 1070° in 1991 with 0.9% CuO (mol%), 0.9% TiO₂, 0.1% B₂O₃, and 0.1% MgO. Full sintering took an hour at 1070°, but only about 15' at 1200°; their result had even higher fracture toughness than undoped alumina. (They also mention as an aside that 1500° to 1700° is adequate to sinter high-purity alumina, and that there’s a known 1096° eutectic between Al₂O₃, CuO, and Cu₂O.)
A perhaps even more effective flux for sapphire may be the MgCl₂-NaCl mixture in US patent 7,988,763, although this is a flux in the metallurgical sense — it fulfills various functions, including scavenging contaminants from the metal and preventing oxidation of the molten metal, but also including liquefying metal oxides. The patent mentions that MgCl₂ mixes with other salts are already in use to keep the melting point of the flux in the 400–500° range. Unfortunately, the patent doesn’t quantify how low it reduces the melting point of the aluminum oxide; it does mention that the mixture was prepared at 550° in an alumina crucible which it apparently did not dissolve in 45 minutes, and that variants of it were used to purify molten aluminum at 850°, 720°, and 700°, but it does not affirm that it liquefied the sapphire dross at these temperatures.
NaCl boils at 1413°, so using it in a pottery kiln salt-glazes both the pottery and the kiln and releases chlorine gas. And that would also be true, and very objectionable, if you heated up sapphire to its normal sintering temperature of 1500–1700°. But if it’s possible to use small amounts of these salts to flux alumina at temperatures more like 400° to 900°, no significant amount of salt should boil off.
By far the most famous flux for sapphire, however, is cryolite (Na₃AlF₆, melting point 1009°) with an excess of aluminum fluoride, which is used in the Hall-Héroult process to dissolve sapphire at 950–980° so that aluminum can be electrolytically produced from it. The binary eutectic melts at 962.7° and contains about 22% alumina, and apparently there is no solid solubility between the compounds, so cryolite may not in fact be very suitable for facilitating the sintering of granular sapphire.
Chromium(III) oxide, the rare mineral eskolaite, also known as chromia, is the surface film that makes chromed bumpers and common stainless steels stainless. It’s a green pigment (“viridian” or “chrome green”) and harder than quartz, though slightly softer than sapphire. It’s sold for pigment use at about US$25/kg.
Viridian is super fucking refractory, not melting until 2435° or sintering until 1600° in pure form, but you can sinter it at 1280° by adding 1% TiO₂ by weight, according to Callister, Johnson, Cutler (the sapphire dude), and Ure 1979. A tricky problem is that under low pressures it tends to smelt to chromium at temperatures above 1600°, and if exposed to oxygen it tends to not sinter, so the sintering has to be done in a non-oxidizing (endogas or argon) atmosphere. The rutile also may serve to stabilize the oxidation state of the viridian.
Viridian is miscible with alumina in all proportions.
According to Nagai and Ohbayashi (1989), viridian is a P-type semiconductor, but the addition of over 2% rutile makes it N-type.
Rutile, titanium dioxide, is commonly used as a white pigment, including in food. In fact, 70% of worldwide pigment production is synthetic rutile. It’s a medium-hard mineral, in the same Mohs 6–6.5 range as orthoclase, a bit softer than quartz. It melts at 1800° and costs about US$20/kg on eBay.
Likely fluxes include oxides of iron (FeO), magnesium, and manganese; more improbable candidates include alkali halides and oxides of zinc and vanadium. The spinel group consists of solid solutions of oxides of titanium, magnesium, zinc, iron, manganese, aluminum, chromium, and silicon crystallized in a cubic close-packed lattice, but titanium spinels are rare, naturally occurring only as ulvite, iron titanium oxide.
Cho and Biswas did find that doping rutile (actually its polymorph anatase) with 1.3% (mol%) vanadium sped up its sintering dramatically, but none of their tests were at a low enough temperature to see if the doping lowered the sintering temperature; their coolest test was at 900°. They do predict theoretically that their “V-TiO₂” should sinter at 800°, while pure TiO₂ basically won’t.
PZT by far the best piezoelectric material, the best high-capacity dielectric, and the best ferroelectric material. It is normally sintered above 1250° but can be sintered at 1000° by doping it with bismuth oxide and carbonates of sodium or lithium (e.g. 0.375% wt% of lithium carbonate and “an equal mole fraction of Bi₂O₃”); this according to Cheng, Fu, and Wei, 1989. Corker, Whatmore, Ringgaard, and Wolny reported in 2000 being able to sinter it at 800° with a “sintering aid” of 5% liquid oxides of lead and copper; the eutectic mixture of PbO (litharge or massicot) and Cu₂O (cuprous oxide or cuprite), which is 80% PbO, melts around 680°, while PbO alone doesn’t melt until 888°, and Cu₂O doesn’t melt until 1235°. (Given PbO’s well-known use as a flux and thinner for glasses and pottery glazes, I suspect this mixture would also work well as a “sintering aid” for quartz, and might result in a transparent article, though Cu₂O alone is so red it’s commonly used as a pigment, and might precipitate out as lead silicate forms.)
Corker et al. also briefly survey the overall use of such sintering-temperature-reducing additives in ceramics, typically metal oxides, citing a pair of papers by Kingery in 1959.
While probably not useful for any practical use, sodium and potassium chlorides may provide an inexpensive way to test and debug apparatus. The NaCl-KCl system has a eutectic point at about 670° at about 50% KCl; NaCl by itself melts at 801°, while KCl by itself melts at about 770°.
This is not an ideal test system from the point of view of the relative concentrations, the relatively low freezing point depression of the eutectic, the hygroscopic and corrosive nature of the powder, and the purely ionic nature of the bonding in the powder; however, we can easily build a “furnace” that reaches 700°, and the two feedstocks are very cheap, nontoxic, and readily available. Potassium chloride costs US$25 for 10 pounds on eBay (US$5.50/kg); on Mercadolibre sodium chloride costs AR$285 (US$18) for 50kg (US$0.36/kg).
According to US Patent 7,988,763, the ternary MgCl₂-KCl-NaCl system has a eutectic point of only 383°, where all three components are equal, and a binary eutectic between MgCl₂ and NaCl at 45% wt% NaCl melting at 439°. Magnesium chloride, which melts at 714°, is even easier to obtain than KCl, and on Mercadolibre it costs AR$400 (US$25) for 5 kg (US$5/kg).
Applying small amounts of mercury to many different metals can produce stable amalgams.
Could you use a powder of some other metal to flux copper powder?
Pure copper, though somewhat expensive, is easily obtained in the modern economy (or from electrical scrap) and does not melt until 1084°; sometimes it occurs naturally allied with arsenic (“bronze”), which both fluxes it so that it can melt at much lower temperatures and hardens it greatly. The humans started using natural bronze some 5300 years ago and synthetic bronze made by mixing separately-occurring tin, rather than arsenic, with copper about 6500 years ago, at first in Thailand and later in the Middle East. Other elements are also widely alloyed with copper, notably aluminum (“aluminum bronze”), zinc (“brass”), zinc and nickel (“nickel silver”), and silicon (“silicon bronze”), with much the same results.
Beryllium forms a precipitation-hardenable alloy with copper that can be harder and stronger than most steels (an ultimate tensile strength of some 1150 MPa at 2% Be), but is little used because of the toxicity of beryllium and beryllia; I do not think it lowers copper’s melting point much in the tiny quantities (0.2%–3%) required to make precipitation-hardenable beryllium copper. XXX yes it fucking does
According to 2001 edition of the ASM Specialty Handbook Copper and Copper Alloys, the solubility limits of these elements in copper at 20° are 6.5% for arsenic, 1.2% for tin, 9.4% for aluminum, 30% for zinc, 0.2% for beryllium, and 2% for silicon, all by weight; it’s miscible in all proportions with gold, nickel, and platinum.
Arsenic is somewhat dangerous, vaporizing at 615° and producing deadly arsenous oxide gas; this may be a significant reason for the switch from arsenical bronze to tin bronze in the early Bronze Age.
Typical bronze is around 12% tin (the rest being copper) and melts around 950°, while tin itself melts at 232°, but remains liquid until 2600°. The tin–copper system is not very friendly for this kind of thing; its eutectic composition is I think Indalloy 244, melting at 227°, which is 0.7% copper, with the rest being tin; the copper is present as a small amount of the brittle intermetallic Cu₆Sn₅ rather than forming an alloy. This suggests that you’d really just be kind of just soldering the copper particles together with the tin, rather than dissolving them fully to make a homogeneous alloy.
Brass — copper fluxed with zinc — melts at 900° to 940°. Zinc melts at 420° and boils at 907°, producing a serious danger of zinc oxide fumes — mixed with air, the zinc vapor burns immediately, and the oxide solidifies as particles with a diameter of around a nanometer, since the oxide is solid until 1975°. So Roman brass was made by the “cementation” process of heating calamine (zinc carbonate or silicate) with copper and charcoal (presumably to scavenge the oxygen), rather than from metallic zinc; or by heating zinc oxide together with copper. This works by producing zinc vapor directly from the calamine, which is taken up by the solid copper, but the reaction is slow; it typically takes several hours. Roman brass is typically 20% to 28% zinc, but some modern brasses have even higher levels, up to 40%.
So, again, the likely outcome of trying to flux copper powder with zinc metal would be just sort of soldering the copper particles together with zinc.
Brasses and bronzes increase in strength as the degree of copper is reduced; when 60% cold-reduced, 8%-tin bronze has an ultimate tensile strength of some 750 MPa, and 30%-zinc brass is nearly as strong at 600 MPa, compared to some 400 MPa for pure copper or 225 MPa for pure annealed copper. But the objects produced by this flux-deposition process would be fully annealed, so their strength would be more like 350 to 400 MPa even at those high alloying levels
