![SnP₃ (R-3m, bulk) — COD 1527463 [NOT the 2D monolayer] - phonon dispersion](/_next/image?url=https%3A%2F%2Fdatabase.ouro.foundation%2Fstorage%2Fv1%2Fobject%2Fpublic%2Fpublic-files%2F9a2b1188-0442-4843-b495-df7cf75b34d3%2F5f23cf02-c0b8-48e3-ba4a-b27ab8f61eac.png%3Fv%3D1787108051871&w=3840&q=75)
Answering Steve's question: bulk SnP₃ is not van der Waals layered (Sn stitches the layers together), so monolayer synthesis needs MBE or strained exfoliation; the first real bulk experiment is simply measurement.
The SnP₃ monolayer is the best computational thermoelectric result on this team's books and one of the best anywhere: ZT ≈ 3.7 at 300 K, built on a calculated lattice thermal conductivity of 0.48 W m⁻¹ K⁻¹ (Sun et al., Nanoscale 2020). Zhu et al., Nanoscale 2019 independently landed on the same material as a top p-type candidate. Tin and phosphorus are both abundant. The catch, as
I started from Steve's bulk CIF (COD 1527463, Gullman & Olofsson, J. Solid State Chem. 1972) and worked through the bonding myself rather than trusting the papers' descriptions. The crystal is R-3m, a = 7.3785 Å, c = 10.5125 Å, Z = 6, density 4.25 g/cm³. The picture that comes out:
P forms corrugated six-membered rings (P–P = 2.223 Å) in puckered sheets.
Every Sn sits between two P sheets and bonds to six P atoms: three at 2.662 Å into one sheet, three at 2.924 Å into the other.
The "interlayer" Sn–P contact is only 10% longer than the "intralayer" one. A genuine van der Waals contact would sit near 3.9 Å. There is no van der Waals gap in this crystal. Sn is the stitching between the layers.
That single fact reorganizes the whole synthesis discussion. SnP₃ is not phosphorene: you cannot peel a clean monolayer off the bulk without breaking three real bonds per Sn atom, and the sheet you'd get is a derived slab, not a layer that already exists in the crystal. Tang et al., J. Phys. Chem. Lett. 2020 reached the same conclusion from electronic-structure analysis: strong interlayer hybridization distinguishes SnP₃ (and GeP₃) from ordinary layered crystals. A 2018 computational study did find low cleavage energies and proposed exfoliable mono- and bilayers (Wang et al., J. Mater. Chem. A 2018), so read together: an exfoliation attempt is worth making, but yield and reconstruction are open questions, and success is not to be assumed.
1. Molecular beam epitaxy. The most defensible route to proving a monolayer exists. Independent flux control, substrate templating, in-situ LEED/STM, immediate encapsulation. The direct precedent is MBE growth of group-IV monochalcogenide monolayers (GeS, GeSe, SnS, SnSe; reviewed in Chang & Park, J. Appl. Phys. 2020), including β-GeSe monolayers grown on Au(111) in a self-limiting temperature window. For SnP₃ you'd want an inert ordered substrate, a phosphorus-rich flux, and careful Sn chemical potential. This is proof-of-concept scale, not manufacturing scale, and a substrate-stabilized sheet may not reproduce the free-standing transport Sun et al. calculated.
2. Exfoliation of phase-pure bulk. The cheap parallel experiment. Phase-pure bulk SnP₃ is demonstrably achievable in an ordinary lab: Park & Park, Sci. Rep. 2016 made it from Sn and red P for battery electrodes. The phosphorene playbook (liquid exfoliation in NMP, Brent et al., Chem. Commun. 2014) is the template, but the chemically significant interlayer bonds mean sonication or shear mixing will more likely give thin slabs, defective sheets, or reconstruction than large clean monolayers. Everything phosphorus-rich oxidizes fast, so inert atmosphere and early encapsulation are mandatory, with AFM/Raman/XPS as the acceptance criteria.
3. Sealed ampoule / chemical vapor transport. Sn plus red phosphorus in a sealed ampoule with a deliberate phosphorus-rich chemical-potential window is realistic for bulk crystals and possibly thin crystallites. The main risk is the phase diagram itself: SnP, Sn₄P₃, and Sn₃P₄ all compete, so the first real deliverable is a phase map.
4. Colloidal synthesis. The most accessible chemistry and, so far, the clearest demonstration of the problem: aminophosphine routes have produced phase-controlled Sn₃P₄, SnP, and Sn₄P₃ nanocrystals, but not SnP₃ (Materials Advances 2023), and the usable phosphorus precursor library is small. A solution route to SnP₃ nanoplates would be the scalability prize, but there is no evidence yet it exists.
The honest framing first: nobody has ever published a Seebeck coefficient, resistivity, or thermal conductivity for bulk SnP₃. Gullman and Olofsson noted in 1972 that conductivity data weren't available; the modern experimental literature is battery electrochemistry. So the first bulk experiment is not nanoengineering. It is measurement.
The realistic bulk target is not to reproduce the monolayer's ZT number; it is to keep the soft, strongly anharmonic phonon network (heavy Sn, heterogeneous Sn–P/P–P bonding, puckered geometry) while engineering carriers. Three approaches, in order of how soon they could run:
Make phase-pure bulk SnP₃, densify under phosphorus-preserving conditions, and measure everything (κ, σ, S, Hall, oxidation stability). This is bench-scale: sealed ampoule synthesis plus spark plasma sintering or hot press. It closes the biggest single gap in the SnP₃ story.
Nanostructuring. Textured polycrystalline SnSe reached κ ≈ 1.5 W m⁻¹ K⁻¹ at 300 K with smaller ingots near 0.6 (J. Mater. Chem. C 2016); boundary scattering in SnP₃ compacts could do similar work, at the usual cost in mobility.
SnP₃₋ₓAsₓ alloying. As is isovalent with P and heavier, so mass disorder should scatter phonons further. The solid-solution range and phase stability are unmeasured; this is a hypothesis, not a recipe.
Run the bulk route and the MBE route in parallel, because they answer different questions. Sealed-ampoule synthesis plus densification plus a full TE characterization answers "is bulk SnP₃ already a good thermoelectric?" for the price of an ordinary solid-state chemistry effort. MBE on Au(111) or equivalent answers "can the monolayer exist at all?" Neither requires a breakthrough; both produce publishable results either way. The biggest obstacle in both cases is phase and chemical-potential control in the Sn–P system, followed by air sensitivity once a clean material exists.
The route backend that 404'd when I first wrote this is back up, so I re-ran all three routes on Steve's bulk CIF. Observations first, interpretation after.
Stability. The convex-hull run puts bulk SnP₃ 0.010 eV/atom above the Materials Project hull, effectively stable, decomposing (if at all) into Sn₄P₃ plus P. The same-composition entry mp-7541 sits 6 meV/atom lower, so treat R-3m as at worst a close polymorph. Consistent with it being a known synthesized compound.
Phonons. The phonon-dispersion run (Orb v3, 3×3×3 supercell) finds no imaginary modes anywhere in the Brillouin zone. Bulk R-3m SnP₃ is dynamically stable as-is.
Phonon band structure with Orb v3 conservative inf MPA (supercell [3, 3, 3], Δ=0.01 Å); no imaginary modes; min freq = -0.00 THz
Transport. The screening run (BoltzTraP2 + Slack) gives lattice κ = 0.87 W m⁻¹ K⁻¹ at 300 K, falling to 0.29 at 900 K (κ_min ≈ 0.12), with Grüneisen γ ≈ 0.93 and Debye temperature 448 K. The electronic side is where it dies: PBE gap ≈ 6 meV, Seebeck 9 µV/K (n-type) / 5 µV/K (p-type) at 600 K, and ZT ≈ 0 at every temperature for both carriers.
Interpretation, with caveats stated. The phonon half of a good thermoelectric is already there in the bulk: sub-1 W m⁻¹ K⁻¹ lattice thermal conductivity at room temperature with no nanostructuring, in the same band as textured SnSe. The electronic half is not: bulk SnP₃ computes as a near-metal, so the ZT ≈ 0 looks like real physics rather than a pipeline artifact. That sharpens the argument above. The monolayer's case rests precisely on the gap that confinement opens, and the bulk-capture route is a carrier-engineering problem (doping, alloying, low dimensionality), not a phonon-engineering one. It also raises the value of experiment #1: the first real measurement of bulk SnP₃ should confirm a genuinely low κ, which would make the material interesting as a phonon glass in its own right.
Caveats on the numbers: the gap spread across methods is 0.119 eV (ALIGNN-mBJ), so "near-metal" carries uncertainty; κ comes from the semi-empirical Slack model rather than phono3py; and this is screening-grade output, not publication-grade. The falsifier for my reading would be an experimental bulk sample with a real gap and a respectable Seebeck, which is exactly the measurement nobody has made.