The question: does In₄BPt₆, last loop's new boride of the double-kagome metal Pt₃In₂, stay stable when we stop taking shortcuts in the calculation, and is there anything that rules out making it?
No new candidate.
This loop tested In₄BPt₆ instead of looking for new cells. It held up. In₄BPt₆ is still one check short of a full claim: every check has passed except a calculation in a larger basis set.
Why:
Relaxation. Relaxing every competing phase in DFT moved its margin from −11.9 to −12.8 meV/atom below the hull. Finer k-point and smearing settings left it there.
Phonons. The DFT cell has no unstable modes.
Arrangement. Boron belongs in the Pt₆ octahedron: every other placement of two borons we tried is at least 21 meV/atom higher.
Makeability. Boron has nowhere better to go among Pt₃In₂'s neighbouring phases. The host itself is the room-temperature phase.
What we know now that we didn't:
Stacking. Boron almost erases the energy difference between two ways of stacking the double kagome layers. A real sample will likely have stacking faults.
Shortcuts. In this chemistry, single-point calculations were already within 1 meV/atom of the fully relaxed answer.
DFT: PBE in ABACUS, DZP basis, k-spacing 0.2 Å⁻¹, non-spin-polarised. Every In–Pt–B phase in the balance (11 cells) was relaxed in cell and positions. Settings checks were single points on the relaxed cells. Screens: Orb v3 on the hull route
Cell | Geometry | PBE vs compound hull (meV/atom) | At k-spacing 0.12 | At smearing 0.02 eV | Imaginary modes | Energy above the P6₃/mmc form (meV/atom) |
|---|---|---|---|---|---|---|
Target: at or below the PBE hull, beyond the calculation's own error.
The balance. In₄BPt₆ is −12.8 meV/atom below 2 InPt + Pt₃In₂ + BPt. It is −12.9 below 4 InPt + BPt₂, so the two decompositions are effectively tied. Last loop's single points gave −11.9.
How far to trust it. We reran published reactions among these phases with the relaxed cells.
The In–Pt reactions come out 7 meV/atom off.
The B–Pt reactions are within 2.
The order of two BPt₂ polymorphs is 19 off. Putting that order right moves the margin to −9.7.
Either way, the margin is larger than the setup's error.
The polymorph. Look at the boron between the two platinum kagome sheets, in its Pt₆ octahedron. This is the DFT-relaxed cell:
Cell + ionic relaxation with ABACUS DFT; 0.04 eV/Å threshold; final energy = -57377.2371 eV; energy change = -0.0645 eV
Arrangement. We placed two borons in every distinct pair of the 98 empty sites of the Pt₃In₂ cell (247 cells) and relaxed them on the screen.
Every boron that did not end up in a Pt₆ octahedron sat at least 21 meV/atom higher.
Six cells came out 0.5 meV/atom below the candidate on the screen. There the bilayers had slid into the rhombohedral stacking of Fe₃Sn₂ (one of them). DFT puts that stacking 0.7 meV/atom above, so the original stacking stays lowest. For the empty Pt₃In₂ host, the same stacking costs 13 meV/atom.
Makeability. We put boron into every hole of the made phases next to Pt₃In₂ (In₃Pt₂, InPt, In₉Pt₁₃), 22 cells. All were at least 142 meV/atom above the hull on the screen. The lowest, boron in In₉Pt₁₃, decomposes into In₄BPt₆ and InPt.
At least one filled cell from a host nobody has filled is new and on the PBE hull: supported, and stronger. In₄BPt₆ kept its margin through relaxation of every phase and through the settings checks, as predicted (−15 to −9 meV/atom).
Every other boron arrangement is higher: supported. We predicted the mixed placements would sit at least 15 meV/atom higher, and they sit at least 21 higher. We did not predict the stacking result: the R-3m cells tie on the screen and are 0.7 meV/atom above in DFT.
In₄BPt₆ is now one check short of a full claim. That check is the basis set.
For this kind of check, DFT single points at the screen's geometry were enough: relaxing every phase moved the answer by 0.9 meV/atom.
An arrangement screen has to look at the symmetry of each relaxed cell, not only its energy. Ours found a second stacking that a screen comparing energies alone would have counted as one more arrangement.
In₄BPt₆ is a calculation, not a synthesis.
Basis set. The margin has not been checked in a larger basis (TZDP).
Synthesis. Pt₃In₂ forms only below 550 °C. The one recent synthesis reacted the elements at 1050 °C, then annealed at 525 °C for six days. A boride would need the same slow route, and how fast boron moves at those temperatures is unmeasured.
Stacking faults. The two stackings are within 1 meV/atom of each other, so a sample may contain faults.
Earlier: loop 1
0 |
+13.4 (vs made Pt₃In₂) |
No new candidate; the loop-1 boride is −12.8 meV/atom below the hull with every phase relaxed, and only a larger-basis check is left.