Magnetic ordering in VI3: a van der Waals material combining vastly different magnetic anisotropies

A. Koliogiorgos; K. Carva; K. K. Pokhrel; N. Machavcova; S. Ray

Magnetic ordering in VI3: a van der Waals material combining vastly different magnetic anisotropies

T0 review · 4 major / 6 minor · reviewed 2026-07-12 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Mixing two vanadium orbital types with opposite interlayer exchange explains why monolayer VI3 orders hotter than the bulk and implies TC can be tuned by their ratio.

desk verdict Solid ASD study that links HO/LO coexistence to the VI3 bulk/ML TC anomaly; the mechanism is plausible, but opposite-sign JL and fitted concentrations are load-bearing and under-constrained. read the letter →

arxiv 2607.02759 v2 pith:DMRNRS22 submitted 2026-07-02 cond-mat.mtrl-sci

K. K. Pokhrel , S. Ray , N. Machavcova , A. Koliogiorgos , K. Carva This is my paper · ORCID

classification cond-mat.mtrl-sci

keywords 2DvdWmaterialsatomisticspindynamicsorbitalconfigurationsmagneticanisotropyVI3CurietemperatureinterlayerexchangeHO/LOcoexistence

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

VI3 is a layered van der Waals magnet that unusually has a higher Curie temperature in the monolayer than in the bulk, opposite to the usual idea that interlayer coupling helps stabilize order. The paper argues that real samples contain two coexisting vanadium electronic configurations—one with high orbital moment and huge magnetic anisotropy, one with low moment and weak anisotropy—whose energies are nearly equal. Atomistic spin-dynamics simulations that assign each type its own anisotropy and exchange parameters show that raising the fraction of high-anisotropy sites hardens the system against thermal spin flips and raises TC. At the same time the two types produce a spatially patchy interlayer super-superexchange network, including pathways of opposite sign, which frustrates coherent bulk order while leaving strong in-plane ferromagnetism intact. When the high-orbital fraction is near one half the model recovers the experimental bulk TC; a higher fraction recovers the monolayer value. The same concentration sensitivity implies that deliberately shifting the balance between the two vanadium types could tune the ordering temperature over a wide range.

What carries the argument

Two-environment layered Heisenberg Hamiltonian (Eq. 1) with site-type-dependent K, J1 and JL (Table 1), evolved by stochastic Landau–Lifshitz–Gilbert dynamics; the HO/LO concentration CHO is the single control parameter that simultaneously sets average anisotropy and the topology of competing interlayer paths.

What would settle it

Measure the HO/LO population ratio independently (e.g., by X-ray absorption or neutron scattering) on the same bulk and monolayer samples used for magnetometry; if the ratio does not track the TC values predicted by the model (near 0.46 for bulk, near 0.76 for monolayer), the concentration-based explanation fails.

Watch

Extended reading notes

Core claim

A random mixture of high-orbital (HO) and low-orbital (LO) vanadium sites, each carrying its DFT- or literature-derived single-ion anisotropy and exchange (including opposite-sign interlayer JL), reproduces the measured bulk Curie temperature near equal populations and the higher monolayer TC at a higher HO fraction; the anisotropy contrast plus the resulting nonuniform, partly competing interlayer network is what both suppresses bulk order relative to the monolayer and makes TC strongly concentration-dependent.

Load-bearing premise

That the two vanadium types really form a random, non-segregated mixture whose local parameters fully capture sample inhomogeneity, and that the only important difference between bulk and monolayer is a shift in that single concentration.

Editorial extensions

If this is right

Bulk and monolayer TC of VI3 can both be recovered from the same microscopic Hamiltonian simply by changing the HO fraction.
Competing interlayer pathways of opposite sign can suppress, rather than reinforce, magnetic order in a van der Waals crystal.
Deliberate control of the HO/LO balance (via doping, polarons, or surface conditions) should allow TC of VI3 to be tuned by at least a factor of two.
The same two-environment picture supplies a microscopic reason why thickness, pressure or stacking changes that alter orbital occupations will move TC strongly.
Related V-trihalides whose orbital configurations are similarly close in energy may show analogous bulk–monolayer anomalies.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If polaron density sets the HO/LO ratio, electrostatic gating or controlled defect introduction becomes a practical knob for TC engineering in VI3 devices.
The opposite-sign JL values suggest that local lattice relaxations around polarons could flip interlayer bonds, offering a real-space picture of how stacking faults or pressure-induced stacking changes reverse interlayer coupling.
Similar random mixtures of high- and low-anisotropy sites may be relevant in other partially filled t2g magnets where orbital degeneracy is nearly unresolved, not only in vanadium trihalides.
A controlled HO-rich surface layer on a LO-rich bulk crystal would be predicted to host a higher-TC skin, testable by surface-sensitive magnetometry.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The new piece here is finite-temperature atomistic spin dynamics on a mixed HO/LO VI3 lattice, not the HO/LO picture itself. They take the large SIA contrast and the two electronic configurations already in the literature, run UppASD, and show that TC rises strongly with HO fraction. The HO+MLH control (same exchange, LO anisotropy only) cleanly separates anisotropy averaging from exchange inhomogeneity. That control is useful and the paper does it carefully.

What they do well: homogeneous K and JL scans first, then the mixed model, susceptibility-peak TC, ensemble averaging over random assignments, and an honest comparison to the Torelli–Olsen 2D formula and Irkhin spin-wave expression. The claim that nonuniform, partly competing interlayer super-superexchange can weaken bulk coherence while leaving intralayer FM intact is physically sensible for a vdW magnet and explains why ML TC can exceed bulk for c_HO ≲ 0.4 under their parameters. Reproducing ~50 K bulk near 1:1 and ~60 K ML near 0.76 is consistent with the spectroscopy/neutron assumptions of ~50%.

Soft spots, in proportion: c_HO is chosen to match the experimental TC for each thickness, so the “agreement” is partly by construction. The opposite-sign JL (HO +0.54, LO −0.38 meV) is the other load-bearing input; the paper itself notes JL is highly sensitive to small distortions and stacking and can flip sign. The LO value comes from a different literature geometry than their HO DFT. If real mixed-network JL values are same-sign or weaker, the bulk–ML inversion shrinks or disappears. The polaron/exfoliation story for the concentration shift is plausible but not measured here. Random non-segregated mixing is assumed, not proven. None of this breaks the paper; it just means the tuning claim and the anomaly explanation are only as secure as those inputs.

This is for people already working on vanadium trihalides or quasi-2D magnets who need a concrete finite-T model of orbital inhomogeneity. Math and methods look standard and solid; citations are appropriate. I would send it to referees. Worth engaging if you care about VI3 or thickness-dependent TC in vdW magnets; not a must-read outside that circle.

Referee Report

4 major / 6 minor

Summary. The manuscript studies finite-temperature magnetism in VI3 using atomistic spin-dynamics (UppASD) on a layered Heisenberg Hamiltonian with single-ion anisotropy, informed by DFT (ELK, GGA+U+SOC) and literature parameters. It models two coexisting V environments—high-orbital-momentum (HO) and low-orbital-momentum (LO)—with strongly contrasting K and environment-dependent J1 and JL (Table 1). Homogeneous scans show TC rising with K (sub-linearly) and with JL. In a random HO/LO mixture, increasing c_HO raises TC; bulk experimental TC (~50 K) is recovered near c_HO ≈ 0.46 and monolayer TC (~60 K) near c_HO ≈ 0.76. An HO+MLH control isolates anisotropy contrast from exchange inhomogeneity. The authors argue that anisotropy contrast plus spatially nonuniform, partly competing interlayer super-superexchange weakens coherent bulk interlayer order while preserving intralayer FM correlations, thereby explaining the anomalous ML > bulk TC and implying that TC can be tuned over a broad range by the HO/LO ratio.

Significance. If the two-environment picture and the opposite-sign JL assignment hold, the work offers a concrete microscopic account of an experimentally reported anomaly (ML TC higher than bulk) that standard homogeneous spin-wave treatments miss, and it links that anomaly to orbital-configuration coexistence already suggested by spectroscopy and neutron work. Strengths include: standard, transparent ASD methodology; explicit control models (homogeneous K and JL scans; HO+MLH anisotropy-only alloy) that separate mechanisms; first-principles JL for both configurations; and a falsifiable prediction that TC is highly sensitive to the HO/LO ratio. The framing that weak interlayer coupling need not always stabilize order when it is spatially competing is of broader interest for vdW magnets. The main limitation is that the quantitative bulk–ML story and the fitted concentrations rest heavily on under-constrained JL values and the random-mixing assumption, so the significance is conditional on those inputs surviving robustness checks.

major comments (4)

Table 1 and §3.3: The central bulk–ML claim relies on opposite-sign effective interlayer couplings (HO JL = +0.54 meV, LO JL = −0.38 meV) together with a random HO/LO mixture (Fig. 7). The manuscript itself notes that JL is highly sensitive to small lattice distortions and stacking and can change sign. The LO JL is taken from a different literature geometry than the HO DFT calculation. Fig. 5b shows ML TC exceeding bulk only for c_HO ≲ 0.4 under the present parameters. Please add a systematic sensitivity analysis: same-sign JL of reduced magnitude, JL scaled by ±50%, and/or JL set equal for HO and LO while keeping K and J1 contrast. Without this, it is unclear whether competing interlayer paths are required or whether anisotropy contrast alone (plus a c_HO shift) can produce ML > bulk.
§3.3, Figs. 5b and 8: Bulk c_HO = 0.46 and ML c_HO ≈ 0.76 are chosen so that simulated TC matches experiment. Agreement with the ~50% assumed in prior spectroscopy/neutron studies is supportive but not an independent determination of the same quantity. The abstract and conclusions currently read as if the model “reproduces” and “supports” coexistence primarily via TC matching. Please reframe clearly: state which quantities are inputs (K, J1, JL, random mixing) versus outputs (TC(c_HO)), report TC over the full c_HO range as the main result, and present the experimental TC match as a consistency check rather than validation of the concentration itself.
§2.2 and §3.3: The two-environment model assumes a random, non-segregated mixture on the V sublattice, while the literature sometimes speaks of “domains.” The HO+MLH control keeps exchange homogeneous (all JL = JL_HO), so it does not test whether LO-specific negative JL is necessary for the bulk suppression relative to ML. Please either (i) simulate a segregated domain morphology at fixed overall c_HO and compare TC to the random case, or (ii) explicitly justify why random mixing is preferred and quantify how domain size would change the bulk–ML relation. Also clarify how mixed HO–LO interlayer bonds are assigned when only two JL values are tabulated (arithmetic average, geometric, or separate DFT for mixed pairs).
§3.3 and Conclusions: Attribution of the bulk–ML TC difference primarily to a shift in c_HO (polarons/exfoliation/surface orbital occupation) is plausible but not uniquely constrained. Stacking faults, strain, and vdW-gap changes upon exfoliation can also alter JL and TC. Please discuss these alternatives quantitatively where possible (e.g., using the homogeneous JL scan of Fig. 4) and state what experimental signature would distinguish a c_HO change from a stacking/gap change.

minor comments (6)

§2.1, Eq. (1): Spins are normalized to unity; state explicitly how physical S (and any g-factor) enter when mapping DFT energies to J and K, so that absolute TC scales can be checked against mean-field and spin-wave estimates in §3.2.
§2.2: Supercell is 30×30×10 with 5 ensemble realizations of random HO/LO assignment. Report the statistical uncertainty on TC more visibly in Figs. 5–6 (error bars) and note finite-size checks if available.
Fig. 3 vs Fig. 2: The Torelli–Olsen comparison is useful; clarify whether the vertical offset is fully accounted for by JL alone or also by further-neighbor intralayer J.
Table 1: Cite the precise sources for each entry (K_HO, K_LO, J1_HO, J1_LO, JL_HO, JL_LO) in the table caption or a footnote for reproducibility.
Typographical/notation: “obatained” (Fig. 4 caption); “Density funtional theory”; inconsistent CHO vs c_HO / C_HO notation; “super-superexchange” hyphenation varies.
Abstract and §1: The claim that ML TC is higher than bulk is experimental input; make sure the theoretical result is phrased as “can reverse the usual bulk–ML ordering for c_HO below ~0.4 under our parameters,” consistent with Fig. 5b.

Circularity Check

2 steps flagged · score 5.0 of 10

c_HO is adjusted to force TC agreement with experiment, then the match is presented as independent model reproduction supporting HO/LO coexistence at ~1:1.

fitted input called prediction [§3.3 (text around Fig. 5b and Fig. 8); Abstract; Conclusions]
"Comparing the bulk TC with the prediction allows us to determine c bulk HO =0.46 within the assumptions of our model. … Agreement with the experimentally observed TC =60 K for a ML is achieved for C ML HO approximately 0.76. … Our model replicates the experimental TC if the ratio between the two V types is close to 1:1, in agreement with the results of two experimental methods. Therefore, the finite temperature magnetization behavior supports the picture of two coexisting V configurations in VI3."

c_HO is the sole free concentration parameter. It is varied until the simulated susceptibility peak coincides with the experimental bulk and monolayer TC values; the resulting numbers (0.46 / 0.76) are then reported as the model “replicating” experiment and as independent support for HO/LO coexistence. The match is therefore obtained by construction of the fit, not by an unconstrained prediction from fixed parameters.
fitted input called prediction [Abstract; §4 Conclusions]
"Our model reproduces experimental TC values when the ratio of the two V types is close to 1:1 … The sensitivity of TC to the HO/LO ratio further suggests that the ordering temperature of VI3 can in principle be tuned over a broad range by controlling the relative occupation of the two vanadium configurations."

The broad-range tuning claim inherits the same fitted concentrations that were chosen to match the two experimental TC points. Once c_HO is free to set any TC inside the computed range, the statement that TC “can be tuned” by that ratio is tautological with the fitting procedure rather than an independent forecast.

full rationale

Core Hamiltonian parameters (K, J1, JL for HO/LO) are taken from independent DFT/literature sources and are not fitted to the target TC values; homogeneous K- and JL-scans and the HO+MLH control are therefore non-circular. The load-bearing claim that the two-environment model “reproduces experimental TC” and thereby “supports the coexistence of two V configurations” rests on freely choosing the single concentration c_HO so that simulated TC equals the experimental bulk (~50 K) and monolayer (~60 K) numbers. That choice yields c_HO≈0.46 (bulk) and ≈0.76 (ML), after which the paper notes agreement with the ~50 % ratio previously assumed in spectroscopy/neutron work. Because the free parameter is tuned to the very observable being “predicted,” the reproduction is statistically forced rather than an a-priori forecast. Self-citations for individual parameters are ordinary and not load-bearing circularity. No self-definitional loop or uniqueness theorem is present. The circularity is therefore partial and confined to the concentration-fitting step that underpins the central experimental-agreement claim.

Assumptions & free parameters 4 free parameters · 6 assumptions · 2 invented entities

The central claim rests on classical Heisenberg+SIA dynamics with two site types whose parameters are taken from DFT/literature, plus a single free concentration per thickness fitted to experimental TC. HO/LO themselves are inherited entities with partial experimental support; the paper does not invent a new particle or force but does introduce an effective random two-environment lattice and a fictitious MLH control species. Load-bearing domain assumptions include transferability of static DFT J/K into classical ASD and random (non-segregated) mixing.

free parameters (4)

c_HO (bulk) = ≈0.46
Relative HO site fraction in bulk; set to ≈0.46 so simulated TC matches experimental bulk ~50 K (§3.3).
c_HO (monolayer) = ≈0.76
HO fraction for ML; set to ≈0.76 to match experimental ML TC ~60 K (§3.3, Fig. 8).
Hubbard U and JH on V 3d = U=4.3 eV, JH=1.1 eV
U=4.3 eV, JH=1.1 eV chosen for DFT (§2.3); affect HO/LO preference and JL.
K_HO and K_LO (adopted) = 15.92 meV / 0.40 meV
SIA taken as 15.92 meV (HO) and 0.40 meV upper limit (LO) from prior ab initio/literature (Table 1); not refit here but strongly drive TC(c_HO).

assumptions (6)

domain assumption Classical stochastic LLG / atomistic Heisenberg dynamics with normalized spins adequately describe finite-T order in VI3.
Standard UppASD framework (§2.2); quantum S=1 and orbital degrees of freedom are not dynamical.
domain assumption Magnetic Hamiltonian is isotropic Heisenberg plus uniaxial single-ion anisotropy with only nearest-layer interlayer terms (Eq. 1).
Layered spin Hamiltonian §2.1; longer-range and anisotropic exchange beyond SIA neglected.
ad hoc to paper HO and LO environments form a random mixture on the V sublattice without spatial segregation into domains.
Explicit modeling choice §3.3, justified by disorder signatures in magnon fits but not proven for real samples.
domain assumption DFT-derived J1 and JL (including opposite signs for HO vs LO) transfer to the classical model for each local environment.
Table 1 and §2.3; JL especially sensitive to stacking/distortion as authors note.
standard math TC is identified with the peak of magnetic susceptibility in finite supercells.
§2.2 practical criterion; Binder cumulant/heat capacity not used.
ad hoc to paper Bulk–ML difference in c_HO can be attributed to polaron/charge or surface-sensitive orbital occupation changes upon exfoliation.
§3.3 and Conclusions; qualitative link to XPS polarons and surface spectroscopy, not measured in the same samples.

invented entities (2)

Two-environment random HO/LO spin lattice (effective orbital-driven inhomogeneity model) independent evidence
purpose: Encode coexistence of two V configurations with distinct K, J1, JL without explicit polarons or lattice relaxation.
Constructed in §2.2–3.3 as an effective description; HO/LO labels preexist, but the random mixed Hamiltonian is the paper’s working entity.
MLH fictitious sites (J = J_HO, K = K_LO)
purpose: Control model isolating anisotropy contrast from exchange inhomogeneity (HO+MLH vs HO+LO).
Introduced only for decomposition in §3.3 / Fig. 6; not claimed as a real species.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Magnetic ordering in VI3: a van der Waals material combining vastly different magnetic anisotropies." pith.science (2026). https://pith.science/paper/DMRNRS22

@misc{pith2026260702759,
  author       = {Pith},
  title        = {Pith review of: Magnetic ordering in VI3: a van der Waals material combining vastly different magnetic anisotropies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DMRNRS22}},
  note         = {Machine review of arXiv:2607.02759}
}

read the original abstract

Among magnetic van der Waals materials, the vanadium trihalide family exhibits unique features. In particular, VI3 contains V atoms of two different types, as two energetically close electronic occupations can coexist in real samples. These types show strikingly different magnetic anisotropy, predicted to differ by more than an order of magnitude. The combination forms a distinctive magnetic system. VI3 also displays an unusual thickness dependence: the monolayer Curie temperature (TC) is higher than that of the bulk, contrary to the expectation that interlayer coupling reinforces magnetic order. Using atomistic spin-dynamics simulations informed by first-principles calculations, we investigate the critical temperature behavior from the combined perspective of single-ion anisotropy and exchange interactions. The strong anisotropy contrast significantly affects thermal stability: increasing the fraction of high-anisotropy sites raises the energy cost of transverse spin fluctuations and increases the ordering temperature. Furthermore, V-atom inhomogeneity makes the interlayer super-superexchange network spatially nonuniform, creating competing exchange pathways. This weakens coherent interlayer order while preserving robust intralayer ferromagnetic correlations, thus modifying the bulk-monolayer TC relation. Our model reproduces experimental TC values when the ratio of the two V types is close to 1:1, in agreement with two experimental methods. This supports the coexistence of two V configurations in VI3. The monolayer TC is reproduced with a slightly modified ratio, possibly linked to polaron concentration. The sensitivity of TC to this ratio suggests that the ordering temperature of VI3 can in principle be tuned over a broad range by controlling the relative occupation of the two vanadium configurations.