{"id":"b3f3738d-b3a0-43ac-b0c7-eb8ed24f9e3a","arxiv_id":"1908.07671","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In the K-Γ-Γ' model for α-RuCl3, the chiral Kitaev spin liquid survives only in a small corner, while new nematic paramagnet phases dominate the intermediate magnetic-field regime.","lead":"This paper uses heavy numerical tensor network calculations to map the magnetic phase diagram of a model for the Kitaev magnet α-RuCl3. It finds the exotic Kitaev spin liquid occupies only a tiny region, with new 'nematic paramagnet' phases dominating the intermediate field window.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The NP1/NP2 phase diagram rests on D=6 simple-update iTPS energy comparisons, and the paper's own convergence checks leave the NP2 boundaries and even its existence insufficiently settled.","rationale":"The reader's weakest assumption identifies the convergence and unbiasedness of the iTPS optimization, which is exactly the load-bearing issue. My reading sharpens it: the paper's own supplementary material documents that NP2 is the least converged phase (D=6 vs D=8 magnetization differences), and that the FM/NP2 energy difference is of order 10^-4, which is at or below the resolution of a simple-update fixed-tau ITE. This is not an external objection to tensor-network methods in general; it is a specific internal consistency problem for the new phase boundaries. The authors deserve credit for reporting initial-state dependence, D-dependence, and energy differences, and their tilted-field continuous-transition analysis is a reasonable attempt to mitigate the bias. However, the main phase diagram is still presented at D=6, and the NP2 phase in particular is not shown to be converged. A higher-D full-update rerun, or an independent cylinder DMRG check at a handful of representative points, would settle whether the NP1/NP2 region is a true ground-state feature or an artifact of the optimization. Since the reader's verdict was already CONDITIONAL, my concern does not move the verdict; it reinforces the conditions already stated.","tokens_in":16640,"tokens_out":3251,"duration_ms":35550,"concrete_test":"Recompute the phase diagram at Gamma' = -0.03 with D=8 and D=10 using a full-environment update or a variational iPEPS optimizer, starting from the same set of initial states, and compare variational energies and order parameters at representative points such as (Gamma/|K|, h) = (0.1, 0.15), (0.2, 0.15), and (0.05, 0.1). If the NP1/NP2 regions shrink by more than about 0.05 in Gamma/|K|, or if a more accurate optimization no longer selects the NP2 state over the polarized or zigzag states by an energy margin larger than 10^-3, the central phase-diagram claim needs to be weakened. As an independent cross-check, run iDMRG or DMRG on cylinders of width 4-6 at those points and compare ground-state energies and local order parameters.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that NP1/NP2 occupy a large intermediate-field window rests on energy comparisons from a simple-update imaginary-time iTPS at D=6, a method the authors themselves describe as easily biased by initial states (Methods; Supplementary Note 2: 'the correct phase boundary and its nature might be concealed by such a biased optimization'). The concern is not merely generic: Supplementary Note 3 reports that the NP2 magnetization is not converged at D=6, with visible D=6 vs D=8 discrepancies, and Supplementary Note 8 reports FM vs NP2 energy differences of order 10^-4, smaller than the expected accuracy of a fixed-tau simple update. Because the NP phases are identified by selecting the lowest-energy state among runs seeded from classical product states, a biased optimizer can create spurious first-order-looking boundaries and stable-looking paramagnets that are not true ground states. If the NP region shrinks or its internal structure changes under more accurate optimization, the paper's main reframing of the field-induced phase diagram loses its basis. The topological reading of the P-NP2 transition is a secondary inference and inherits the same convergence issue.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies the K-Γ-Γ' model of Kitaev magnets in a [111] magnetic field using infinite tensor product states (iTPS) with imaginary-time evolution and a simple update. The authors report a phase diagram with a small chiral Kitaev spin liquid (KSL) pocket, a polarized phase, two nematic paramagnet phases (NP1 and NP2) occupying a large intermediate-field window, and a low-field zig-zag phase. They also consider the Γ'=0 K-Γ model and tilted fields, and argue that a continuous transition between the polarized and NP2 phases in a tilted field implies that NP2 is a topologically non-trivial state. The paper includes extensive supplementary material on initial-state dependence, bond-dimension dependence, and energy comparisons.","tokens_in":16968,"tokens_out":4964,"duration_ms":123813,"significance":"The interpretation of the field-induced phase in α-RuCl3 is an important open problem, and the claim that the KSL is confined to a small corner while nematic paramagnets dominate the intermediate-field regime directly challenges earlier finite-size ED and DMRG results. If established, the predictions of C3-breaking anisotropy in thermal conductivity and susceptibility would be of considerable experimental interest. The paper's strengths are its thermodynamic-limit approach, the use of multiple independent initial states, and the supplementary documentation of convergence checks. However, the main claims are not yet supported to the required standard: the central phase diagram depends on D=6 simple-update iTPS, which the authors themselves describe as easily biased, and the convergence of the NP2 phase is not demonstrated.","major_comments":[{"comment":"The phase diagram in Fig. 1 is determined by comparing variational energies from D=6 simple-update iTPS runs, but the Methods section states that this optimization 'can be easily biased by the initial choice of Ti', and Supplementary Note 2 warns that 'the correct phase boundary and its nature might be concealed by such a biased optimization'. As the NP1 and NP2 phases are identified as the lowest-energy states among runs seeded from classical product states, the possibility remains that these phases are artifacts of the biased optimization. Please provide quantitative evidence, such as energy differences relative to estimated error bars or D=8/D=10 sweeps across the full parameter range, that the NP1/NP2 window is stable.","section":"Methods; Fig. 1"},{"comment":"Supplementary Note 3 reports that the NP2 magnetization is not converged at D=6, with discrepancies between D=6 and D=8, and that D=4 does not represent the NP2 state well. Since the NP2 phase is a central finding, the claim that the phase diagram 'will be more or less the same' at larger D is an assertion rather than a demonstrated result. Please show the D-dependence of NP2 energies, magnetizations, and phase boundaries for the parameter range of interest, and specify the accuracy required to distinguish the competing states.","section":"Supplementary Note 3"},{"comment":"Supplementary Note 8 shows that the energy difference between FM and NP2 states is ~ O(10^-4) at Γ/|K| = 0.03 and 0.095. This is below the expected accuracy of a fixed-τ (τ=0.01) simple-update imaginary-time evolution, so the ground-state selection between these phases at low field is not established. Please specify the accuracy of the energy evaluation and explain why energy differences of this magnitude are meaningful despite the acknowledged optimization bias.","section":"Supplementary Note 8"},{"comment":"The conclusion that the continuous P–NP2 transition in a tilted field (θ=5°) implies a topological transition is not fully supported. A continuous transition between two states that break no symmetry could indicate a topological distinction, but it could also reflect other non-symmetry-breaking order or a crossover sharpened by the variational bias. Moreover, the entanglement entropy data in Fig. 3 and Supplementary Fig. 3 are obtained for small cylinder circumference Ly=2; please provide evidence that the EE peak sharpens with Ly and clarify what topological invariant or ground-state degeneracy would distinguish the NP2 phase from a trivial paramagnet.","section":"Fig. 3; Conclusion"}],"minor_comments":[{"comment":"In Eq. (1), the Γ' term appears to contain a typo: the final term is written twice as Sν_i Sγ_j; it should likely be Sγ_i Sν_j. Please correct the expression.","section":"Eq. (1)"},{"comment":"The phase diagram uses the label 'K' for the Kitaev spin liquid phase while 'KSL' is used elsewhere in the text; please unify the notation for clarity.","section":"Fig. 1"},{"comment":"The caption of Fig. 2 does not specify the meaning of the green dotted lines or the shaded regions in the upper panels; please expand the caption to state what quantities are shown and how the phase boundaries are determined.","section":"Fig. 2 caption"},{"comment":"Supplementary Note 3 refers to 'Fig. 2 in the main text' when presumably referring to the phase diagram; this cross-reference should be corrected to the appropriate figure.","section":"Supplementary Note 3"},{"comment":"The statements that code and data are 'available from the authors upon reasonable request' are less reproducible than a public repository deposit; please consider providing permanent access to the exact codes and generated data.","section":"Data and code availability"}],"recommendation":"major_revision","confidential_remarks":"The paper addresses a timely and important question, and the central claim of a small KSL pocket and large nematic paramagnet region is potentially significant. However, the evidence rests on a D=6 simple-update iTPS approximation whose limitations the authors candidly acknowledge. The supplementary notes contain explicit statements that the optimization can be biased and that NP2 is not converged at D=6. In my view, the paper needs a substantial revision to either demonstrate stability of the NP phases and the P–NP2 transition under more accurate optimization, or to soften the claims accordingly. The topological interpretation is a secondary inference that inherits the same convergence concerns."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth reading carefully. The new thing here is a thermodynamic-limit iTPS phase diagram of the K-Gamma-Gamma' model at Gamma'=-0.03: the chiral Kitaev spin liquid survives only in a small corner, and two nematic paramagnets (NP1, NP2) occupy a large intermediate-field band between the zig-zag and polarized states. If true, that reshapes the alpha-RuCl3 debate, because the field-induced phase would not be the chiral KSL. The paper is also honest: the Methods and Supplementary Note 2 state that the simple-update imaginary-time evolution can be easily biased by the initial state and that the correct phase boundary and its nature might be concealed. That kind of candor is rare and it lets a referee focus.\n\nWhat is genuinely good: they compare many initial states (string gas, several FM directions, zigzag, 6/8/18-site orders) and select the lowest energy; they check D=4,6,8 and show D-dependence; tilting the field stabilizes the optimization; the entanglement entropy peak sharpens with bond dimension; and the D-to-infinity extrapolation for the NP magnetization at zero field is reasonable evidence that the NP phases are paramagnets. This goes beyond earlier ED and ladder DMRG work and makes contact with the classical large-unit-cell orders. The citation pattern is appropriate: the string-gas seed and the classical states come from the authors' own prior work, which is normal and not a smoke screen.\n\nThe soft spots are real, and one is load-bearing. First, the KSL phase is found only from the string-gas seed; other initial states do not land there. The authors say this explicitly, meaning the small KSL corner could be a basin artifact. Second, NP2 is not converged at D=6: Supplementary Note 3 shows visible D=6 versus D=8 magnetization differences, and Supplementary Note 8 gives FM versus NP2 energy differences of order 10^-4, comparable to the expected accuracy of a fixed-tau simple update. Since the NP2 phase is defined by energy comparison, its boundaries, and at some parameters its existence, are not settled. That is a serious weakness, not a nitpick. Third, the main text's line that a continuous P-NP2 transition in a tilted field ‘implies a topological phase transition’ overreaches: when the tilted field explicitly breaks C3 in both phases, a continuous transition without a symmetry change can be ordinary. The data show a sharp feature, but not topology. The conclusion hedges with ‘possibility,’ yet the earlier wording is stronger than the evidence. Finally, Ref. 20 already reported a rotation-symmetry-broken spin liquid, so the nematic states are not brand-new phenomenology; the new contribution is the thermodynamic-limit map.\n\nThis paper is for the Kitaev materials community, especially those working on alpha-RuCl3 and on tensor-network phase diagrams. It deserves a serious referee. I would send it out rather than desk reject, but the revision needs to either raise the bond dimension for the KSL and NP2 boundaries or soften the claims about the small KSL corner and the topological transition. I would cite it as the iTPS thermodynamic-limit result, with convergence caveats noted.","headline":"Thermodynamic-limit iTPS study that shrinks the chiral Kitaev spin liquid to a small corner and puts two nematic paramagnets in the intermediate-field window; the NP2 boundaries and the topological interpretation are not yet solid.","tokens_in":17414,"tokens_out":3099,"would_cite":true,"duration_ms":36092,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"In the K-Γ-Γ′ model of Kitaev magnets, the chiral Kitaev spin liquid occupies only a small corner of the magnetic-field phase diagram, while two nematic paramagnetic phases fill the intermediate-field window between zig-zag order and the…","keywords":["Kitaev spin liquid","chiral spin liquid","alpha-RuCl3","nematic paramagnet","tensor network","infinite tensor product states","K-Gamma-Gamma' model","magnetic field induced phases"],"falsifier":"At $h = 0.15$ with the field tilted by $\\theta = 5^\\circ$, the paper places the polarized-to-NP2 transition at $\\Gamma/|K| \\approx 0.05$, signaled by a peak in the second derivative of the energy and in the first derivative of the entanglement entropy; rerunning the optimization at $D = 8$, 10, and 12 and finding that these peaks wash out, shift discontinuously, or vanish—or that the FM[100] and FM[011] initial states converge to the polarized state throughout this window—would show the NP phase and its continuous transition are ansatz artifacts. Experimentally, measuring longitudinal thermal conductivity and magnetic susceptibility along different in-plane directions in $\\alpha$-RuCl$_3$ and finding no $C_3$-breaking anisotropy in the intermediate-field phase would rule out the nematic paramagnets in the material.","tokens_in":16493,"feed_emoji":"🧲","tokens_out":12498,"duration_ms":586821,"temperature":0.7,"pith_summary":"The paper asks which quantum ground states actually appear between the low-field zig-zag magnetic order and the high-field polarized state in minimal models of the Kitaev magnet $\\alpha$-RuCl$_3$. Working directly in the two-dimensional thermodynamic limit with infinite tensor product states, it claims that the chiral Kitaev spin liquid, previously thought to occupy a large window in magnetic field, survives only in a small corner of the phase diagram. Instead, two nematic paramagnet phases, NP1 and NP2, occupy the intermediate-field window; both spontaneously break the lattice threefold rotational symmetry down to a twofold one. If the phase diagram is correct, the field-induced intermediate phase seen in $\\alpha$-RuCl$_3$ is more plausibly a broken-symmetry paramagnet than a chiral spin liquid, and experiments should probe rotational-symmetry breaking rather than Majorana edge signatures alone.","feed_headline":"Tensor networks shrink the Kitaev spin liquid to a corner","feed_subtitle":"Two nematic paramagnets, not the chiral spin liquid, fill the intermediate-field window in the K-Γ-Γ′ model.","key_machinery":"The central object is the infinite tensor product state (iTPS), a variational tensor-network wavefunction on the infinite honeycomb lattice with virtual bond dimension $D$, optimized by imaginary time evolution with a simple local update. The argument is carried by comparing the energies of states evolved from many initial configurations—the string-gas representation of the Kitaev spin liquid, ferromagnetic states along different axes, and the 6-site and 18-site classical magnetic orders—and by characterizing the winner through the flux operator $\\hat{W}_p$, the magnetization components, and the entanglement entropy computed from the tensor-network boundary theory. The tilted-field setup breaks the rotational symmetry explicitly and removes the initial-state bias near the P–NP transition, which is what lets the paper see a continuous transition rather than a first-order one.","core_discovery":"Using the infinite tensor product state (iTPS) ansatz with bond dimension $D=6$, the study maps the ground-state phase diagram of the K-Γ-Γ′ model with ferromagnetic Kitaev coupling, $\\Gamma' = -0.03$, and a magnetic field along [111]. The central discovery is that the chiral Kitaev spin liquid is confined to a small corner of the $(\\Gamma/|K|, h)$ plane, in contrast to the large spin-liquid window reported by 24-site exact diagonalization and two-leg ladder DMRG. The intermediate-field region between the zig-zag and polarized phases is instead occupied by two nematic paramagnetic phases, NP1 and NP2, which break the rotational symmetry of the honeycomb lattice down to $C_2$ and acquire finite magnetization only under the field. A slightly tilted field reveals a continuous transition between the polarized phase and NP2, signaled by a peak in the first derivative of the entanglement entropy; since no conventional symmetry distinguishes these phases, the authors conclude that NP2 is not a trivial product state and may be non-trivial or topological. The same NP phases appear in the K-Γ model ($\\Gamma' = 0$), where they survive down to almost zero field and give way to 6-site and 18-site magnetic orders at larger $\\Gamma$.","pith_inferences":["If the NP phases are indeed non-trivial topological states, the half-quantized thermal Hall signal in $\\alpha$-RuCl$_3$ could in part come from the nematic paramagnets, so future experiments should map the thermal Hall signal together with a rotational-symmetry-breaking probe across the same field window.","The authors' own warning that simple-update imaginary time evolution is initial-state biased suggests the phase diagram should be rechecked with a fully unbiased variational optimization; a natural test is whether the NP1/NP2 regions survive at substantially larger bond dimension or with gradient-based tensor updates.","Because the NP phases appear in both the K-Γ-Γ′ and K-Γ models, other spin-orbital-entangled honeycomb magnets with sizable Kitaev and Γ interactions may exhibit the same nematic paramagnetic window; the paper's in-plane anisotropy prediction gives a concrete signature to look for before magnetization saturates."],"forward_implications":["In the K-Γ-Γ′ model with $\\Gamma' = -0.03$, the chiral Kitaev spin liquid is confined to a small corner of the phase diagram, contradicting the much wider spin-liquid window found by 24-site exact diagonalization and two-leg ladder DMRG.","The NP1 and NP2 phases break the threefold lattice rotational symmetry, so longitudinal thermal conductivity and magnetic susceptibility measured along different in-plane directions should show a $C_3$-breaking anisotropy if these phases occur in $\\alpha$-RuCl$_3$.","Because the polarized-to-NP2 transition is continuous in a tilted field and no symmetry separates the two phases, the NP states are not smoothly connected to a trivial product state and may be non-trivial topological states.","In the K-Γ model ($\\Gamma' = 0$), the NP phases appear even at almost zero field, and the complex 6-site and 18-site magnetic orders from the classical phase diagram re-emerge only for larger $\\Gamma$.","The $\\Gamma'$ interaction weakens the Kitaev spin liquid against the field: the KSL-to-polarized critical field drops from roughly 0.01925 at $\\Gamma' = 0$ to about 0.01075 at $\\Gamma' = -0.03$."],"supporting_citations":[{"why":"The string-gas tensor-network representation of the gapless Kitaev spin liquid; the KSL phase in this study is obtained only from this initial state, so this reference carries the small-KSL-corner result.","marker":"Ref. 1"},{"why":"Supplies the classical phase diagram and the 6-site and 18-site magnetic orders used as initial states and as the reference whose large-unit-cell orders quantum fluctuations melt into the NP phases.","marker":"Ref. 2"},{"why":"Defines the exactly solvable Kitaev model and the plaquette flux operator used to identify the chiral Kitaev spin liquid.","marker":"Ref. 25"},{"why":"The earlier 24-site exact-diagonalization and two-leg ladder DMRG study that reported a large field window for the Kitaev spin liquid; the paper's central disagreement is with this result.","marker":"Ref. 27"},{"why":"The imaginary-time-evolution update scheme used to optimize the tensor network states throughout the study.","marker":"Ref. 30"},{"why":"Previous iDMRG report of a rotation-symmetry-broken spin liquid in similar Kitaev-like models, which the NP phases are said to resemble.","marker":"Ref. 20"},{"why":"Provides the boundary-theory method for computing entanglement entropy from tensor network states; the derivative of this entropy locates the continuous polarized-to-NP2 transition.","marker":"Ref. 39"},{"why":"The half-quantized thermal Hall conductivity measurement in $\\alpha$-RuCl$_3$ that motivates the search for field-induced quantum phases between zig-zag and polarized states.","marker":"Ref. 3"}],"fun_headline_variants":["Kitaev spin liquid squeezed into a tiny corner","Nematic phases, not spin liquid, dominate field window","Tensor networks find two nematic phases in Kitaev magnet","Chiral spin liquid cornered by nematic paramagnets","Nematic paramagnets edge out spin liquid in Kitaev magnet"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that the imaginary-time tensor-network optimization at bond dimension 6, started from the ten chosen initial states, finds the true ground state everywhere in the phase diagram; the authors themselves state that this simple-update scheme 'can be easily biased by the initial choice,' so if the small KSL corner and the NP phases are artifacts of that bias, the central claim collapses.","fun_headline_variants_meta":{"raw":{"variants":["Kitaev spin liquid squeezed into a tiny corner","Nematic phases, not spin liquid, dominate field window","Tensor networks find two nematic phases in Kitaev magnet","Chiral spin liquid cornered by nematic paramagnets","Nematic paramagnets edge out spin liquid in Kitaev magnet"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000495,"raw_usage":{"total_tokens":2434,"prompt_tokens":956,"completion_tokens":1478,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":572,"completion_tokens_details":{"reasoning_tokens":1396}},"tokens_in":572,"tokens_out":1478,"duration_ms":160174,"temperature":1.0,"reasoning_tokens":1396,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:59:36.416964+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"At $h = 0.15$ with the field tilted by $\\theta = 5^\\circ$, the paper places the polarized-to-NP2 transition at $\\Gamma/|K| \\approx 0.05$, signaled by a peak in the second derivative of the energy and in the first derivative of the entanglement entropy; rerunning the optimization at $D = 8$, 10, and 12 and finding that these peaks wash out, shift discontinuously, or vanish—or that the FM[100] and FM[011] initial states converge to the polarized state throughout this window—would show the NP phase and its continuous transition are ansatz artifacts. Experimentally, measuring longitudinal thermal conductivity and magnetic susceptibility along different in-plane directions in $\\alpha$-RuCl$_3$ and finding no $C_3$-breaking anisotropy in the intermediate-field phase would rule out the nematic paramagnets in the material.","supporting_citations":[],"review_version":1}