As I was adding materials data to Likely-feasible non-flux-deposition powder-bed 3-D printing processes I happened across a material I'd never heard of before that seemed like a better fit for what I was looking for than any of the materials I had heard of: aluminum orthophosphate, the mineral berlinite.
I looked into its synthesis, and what I found was absolutely astounding; I must be missing something important, because it seems like a fairly revolutionary new material, but no revolution has resulted in the last 140 years since its discovery.
Berlinite is a rare quartz-like crystal of aluminum orthophosphate with a Mohs hardness of 6.5 and a melting point of 1800°. It's nontoxic enough to be used as an antacid, although there are several other phosphates of aluminum, which are caustic, and a highly toxic phosphide of aluminum, and sometimes these chemicals are confused with aluminum orthophosphate; every MSDS I can find states that aluminum orthophosphate causes severe skin burns, presumably as a result of this confusion; although I don't have any here to put on my skin, I think this is very unlikely to be correct. Its dihydrate is the uncommon turquoise-like mineral variscite, which has a Mohs hardness of only 4.5. It dunts like quartz, but at 583°. It wasn't discovered until 1868.
The really interesting thing about Berlinite is that it's almost as hard as quartz, but several investigators report easy ways to synthesize it at much lower temperatures.
In 1999 some researchers at Argonne National Lab and Purdue investigating the problem of safe nuclear waste storage published a paper on a berlinite-bonded alumina ceramic, entitled "Low-temperature synthesis of berlinite-bonded alumina ceramics". Their stunning claim was that they had synthesized berlinite at 150° by heating a hydrated aluminum phosphate they made by partly dissolving alumina in aqueous phosphoric acid at 130°.
In detail, they heated a slurry of alumina (micron-sized mixed with sand) with 50 wt.% phosphoric acid with an Al2O3:H3PO4 weight ratio of 5:1, detected an endothermic reaction at 118° with differential thermal analysis, held it at 130° for 1, 2, or 4 days to allow the reaction to complete, identified with X-ray diffraction that it produced an intermediate phase of another aluminum phosphate AlH3(PO4)2.H2O mixed with the remaining alumina; they described the mixture as "a thick puttylike...gel", which they left at ambient temperature for another week "so that some crystalline growth would occur". They dried the samples into "hard monoliths", which "disintegrated when placed in water".
All of the above had been done by previous researchers. But then they baked the "monoliths" at 150° for one, two, and three days: "Significant porosity developed in the monoliths as bound water escaped through an increasingly viscous slurry." They infer that the hot monoaluminum phosphate reacted with more of the remaining alumina to produce berlinite over the course of these days, and explain, "the resultant ceramic appeared to be a very hard monolith with dense phases separated by large pores," with some solid phosphoric acid on top; they measured a compressive strength of 6824 psi (47.05 MPa in modern units) on materials with "20.9 vol.% open porosity". This is a bit stronger than ordinary concrete but not remarkably so, and much lower than high-performance concrete.
Mysteriously, this paper has been almost entirely ignored. It's been cited in 2019 by US patent 10,233,803B2 on exhaust-gas filters with a catalyst film on a porous ceramic support (and its related 2017 applications in .de, .br, .uk, .cn, .kr, the WIPO PCT ("WO") and the EPO ("EP")); in 2006 by US patent 6,858,174B2 on gel-casting ceramic slurries; by Stefania CASSIANO GASPAR's dissertation in 2013 at INSA-Lyon on extruding porous berlinite-bonded alumina ceramic supports for catalyst films (in French) for exhaust-gas filtering; by her 2012 patent (US 9,227,873B2) with four other co-authors on the same subject; and by a 2015 Russian paper.
Cassiano Gaspar's dissertation also cites some 2010 work by Lee et al. with zeolites in aluminum phosphate, which she reports they got to be nonporous.
So, why do I think this is such an interesting result? It gives a simple, low-temperature recipe for a moldable porous material that costs about US$3/kg and produce a result as strong as ordinary concrete, which costs a tenth of that. What's so special about that? After all, if you want concrete with alumina abrasives in it, you can put some alumina in your concrete.
Well, a couple of things. First, this material is highly refractory, despite its low-temperature preparation; it will not spall from heat shocks. Second, I think it's nontoxic and noncaustic, unlike portland cement. It might even be biocompatible, which conventional phyllosilicate ceramics are not; alumina is well-known to be biocompatible, phosphate as such is nontoxic, and so the potential biocompatibility concern would mostly be with whether it releases aluminum into the body, catalyzes harmful reactions, or provokes inflammation.
In addition to its use for making massive objects, the resulting berlinite may work well as a binder for connecting aluminum parts together or for a mineral paint similar to those made from waterglass.
Apparently without citing Grover et al., monoaluminum phosphate is sold as a castable refractory, sometimes using the heinous abbreviation "MALP" (to distinguish it from "MAP", monoaluminum phosphate). Fosbind is one brand. Luz, Oliveira, Gomes, and Pandolfelli reviewed its properties, calling it MAP, in a 2016 paper, using 200-micron dead-burnt magnesia and a s00per seekrit 31337 boron source to get it to set in 90-120 minutes, finding that they would work well up to 1400°-1500°; they cite a book (Technology of Monolithic Refractories, Nishikawa) and two papers on the subject in 1982-1989, and point out that it's easier to get aluminum phosphates if your aluminum source is aluminum hydroxide instead of alumina.
Interestingly, Luz et al. give a different formula for "monoaluminum phosphate": Al(H2PO4)3, which is presumably dramatically more acidic than the AlH3(PO4)2 Grover et al. observed, since its phosphate groups still have their second hydrogen. Also Luz et al. experimented with binder systems including Al2O3/H3PO4 in 2015, and found that the most effective recipe was 48 wt.% phosphoric acid with the hydroxide. They claim that their X-ray diffraction results show that the aluminum orthophosphate was hydrated (i.e., variscite, not berlinite). For different compositions, they report flexural strengths in the 10 to 30 MPa range; presumably the compressive strength is much greater.
It would be interesting if silica can somehow be phosphated the same way; normally the answer is no, and phosphoric-acid etching is a normal way to get silicon nitride off of silicon dioxide. But silicon orthophosphate, Si3PO4, does exist. Apparently it's caustic, which suggests that it isn't waterproof, which makes it immediately much less interesting.
Often the reason for wanting to use silica rather than alumina is that alumina costs 300 times as much as silica (see Likely-feasible non-flux-deposition powder-bed 3-D printing processes for materials pricing). But in this case that reason is not nearly as overriding, because phosphoric acid is nearly as expensive as alumina.
The inexpensive 3-D printers that have become popular recently, descended from the RepRap project, are "fused deposition modeling" printers: they feed a filament into a melting chamber to produce pressure to force it out a small heated opening (the chamber and the opening are collectively the "hotend"), moving the opening and the workpiece relative to each other to deposit material in controlled positions on the object produced. Typically they use low-melting polylactide plastic at temperatures of 185° to 225°. Although higher temperatures are sometimes used, most organic chemicals start to break down into simpler substances in the 230° to 270° range. Other popular plastics include PETG, ABS, and nylon (PA6 I think, maybe 6,6.)
There are variants of this process that work well for clay bodies, clay slips, and adobe, largely relying on thixotropic or plastic yielding of the material with some kind of extrusion screw rather than melting. Instead of stressing the "hotend" with heat, these variants instead abrade it.
This approach should be applicable to the pasty aluminum phosphate Grover described, but it is somewhat complicated. If you use alumina as the aluminum source, the alumina particles are highly abrasive; only a very small number of coatings (diamond, CBN, and maybe zirconia, carborundum, and tungsten carbide) can resist being ground away by them, and frequent replacement of consumable liners and nozzles may be the only viable alternative. Also, the aluminum phosphates other than the orthophosphate are highly acidic and consequently corrosive.