{"id":"a4807971-9adc-4141-adcf-6bb709ba3ca8","arxiv_id":"2501.07541","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"At 1.5 electrons per site, the two-dimensional periodic Anderson model hosts a diagonal antiferromagnetic stripe ground state, computed with infinite projected entangled-pair states.","lead":"A numerical study of the periodic Anderson model, a standard model for heavy-fermion materials, finds that at 1.5 electrons per site the ground state is a diagonal antiferromagnetic stripe phase. The result differs from earlier mean-field and variational predictions and shows that larger simulation unit cells are needed for this regime.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim rests on 2x2 supercell energy comparison; larger-supercell runs lack reported energies, so a vertical or longer-period stripe cannot be excluded.","rationale":"The reader's weakest assumption is exactly the one I consider load-bearing. The paper's benchmark at half-filling against QMC is convincing and supports the iPEPS implementation. The n=1.5 claim, however, depends on a small extrapolated energy difference in a 2x2 supercell, and the larger-supercell runs are not reported with energies. The paper's own mean-field result predicting a vertical stripe makes the absence of a systematic larger-supercell energy comparison especially concerning. I also note that for V<0.8 the paper acknowledges overlapping error bars, so the ground-state assignment is not definitive even at the 2x2 level. The proposed test—larger-supercell iPEPS with candidate-order seeding and explicit energy comparison—would settle whether the diagonal AF stripe survives as the global ground state. If it does, the paper's central claim is supported; if not, the claim would need revision. This does not change the reader's conditional verdict; it reinforces the need for the requested verification.","tokens_in":13122,"tokens_out":4808,"duration_ms":45436,"concrete_test":"Run iPEPS with 4x4 and 6x6 supercells at Uf=4, V=1, epsilon_f=-2, n=1.5, initializing with vertical stripe, horizontal stripe, diagonal stripe (period 2 and 4), and uniform ferromagnetic states, and report extrapolated energies per site using the same cost-function extrapolation. If any state other than the 2x2 diagonal AF is lower by more than the combined extrapolation error, the central claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim at n=1.5 (Sec. III B) is that a diagonal AF stripe is the ground state. This is based on an energy extrapolation within a 2x2 supercell (E_AF=-3.766(7) vs E_F=-3.751(4), difference ~0.015). Larger 2x4 and 4x4 supercell calculations are described only as 'exploratory' and no energies are reported; the text states only that the diagonal stripe pattern appeared. This leaves open the possibility that a vertical stripe (which the paper's own mean-field theory predicts) or a longer-period diagonal stripe has lower energy. The small energy difference (about 2 sigma) and the paper's admission that error bars overlap for V<0.8 further weaken the ground-state assignment. Because iPEPS is variational and can miss states not well-represented by the supercell or not reached from the random initial states, the absence of a systematic search over candidate orders in larger supercells is the most load-bearing gap.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents infinite projected entangled-pair state (iPEPS) calculations for the periodic Anderson model on the square lattice. At half-filling and the symmetric point (εf = -Uf/2, Uf=4), the method reproduces the quantum Monte Carlo result for the critical hybridization Vc ≈ 1.2 (QMC: 1.1(1)) and the f-electron magnetization at V=1. The main new claim is that at n=1.5 electrons per site, for Uf=4, εf=-2, V=1, the ground state is a diagonal antiferromagnetic stripe phase, which is lower in energy than a competing diagonal ferromagnetic stripe state (E_AF=-3.766(7) vs E_F=-3.751(4) after extrapolation). The paper also reports that mean-field theory predicts a vertical stripe, not found in iPEPS, and that the diagonal stripe pattern persists in exploratory 2×4 and 4×4 unit-cell simulations.","tokens_in":13308,"tokens_out":10669,"duration_ms":92051,"significance":"If the n=1.5 ground-state claim holds, this is a genuinely new result: it identifies a stripe phase in a heavy-fermion model where previous mean-field, DMFT, and variational Monte Carlo studies predicted uniform ferromagnetism, and it provides a concrete target for future DMRG or large-cell tensor-network studies. The half-filling benchmark is a strong point: the iPEPS estimate of Vc is in excellent agreement with sign-problem-free QMC, and the magnetization at V=1 matches QMC within error bars. The paper also makes a clear falsifiable prediction (the diagonal stripe order) and provides data availability via Zenodo. However, the evidence for the central claim is currently marginal, resting on a 2×2 supercell energy extrapolation with a small energy difference and on qualitative rather than quantitative larger-supercell checks.","major_comments":[{"comment":"The identification of the diagonal AF stripe as the ground state at n=1.5 rests on the extrapolated energies E_AF=-3.766(7) and E_F=-3.751(4) from a 2×2 supercell. The energy difference is only about 1.9σ given the reported errors, and the paper itself states that for V≲0.8 the error bars overlap. Thus the central claim is not robustly established for the parameter range as stated. Please provide a more detailed extrapolation analysis (e.g., alternative fit forms and error estimates) and state explicitly the parameter window in which the AF-F distinction is significant.","section":"Section III B, Fig. 4"},{"comment":"The larger-supercell (2×4 and 4×4) simulations are described only qualitatively ('the diagonal stripe pattern appeared'), and no energies are reported. Since iPEPS is variational and the 2×2 energy comparison is marginal, the possibility that a vertical stripe (predicted by the mean-field calculation) or a longer-period stripe has lower energy is not excluded. To support the central claim, please report the lowest variational energies found in these larger supercells (or provide them as supplementary material) and show that they are consistent with the 2×2 extrapolated value.","section":"Section III B, paragraphs after Fig. 5"},{"comment":"The use of U(1) spin symmetry to target the Stot_z=0 and Stot_z=1 sectors of the 2×2 supercell restricts the search to collinear spin states with fixed total magnetization. The paper's statement that iPEPS provides 'unbiased' results (Introduction and Conclusions) is therefore too strong; the search is unbiased only within a limited manifold. Although the later charge-only simulations partially address this, no energies from those simulations are reported. Please either justify that non-collinear or incommensurate states are not relevant here, or present the corresponding energy data.","section":"Section II and III B"}],"minor_comments":[{"comment":"The definition of filling as 'the ratio of the total electron density per site (n = nc + nf ) to the maximum possible electron occupancy (nmax = 4)' is confusing because n is already a density. Suggest calling n the total electron density and noting that half filling corresponds to n=2.","section":"Section II"},{"comment":"There is a typo in 'techniques that do no rely on rigid Ansätze' (should be 'do not rely'). Also, iPEPS is itself a variational ansatz, so 'no rigid Ansätze' is overstated; consider rephrasing to 'no a priori magnetic-ordering Ansatz.'","section":"Introduction"},{"comment":"The caption says the energy is plotted 'as a function of the iPEPS cost function, w, ... and as a function of the inverse bond dimension,' which suggests two abscissas in one figure. Please clarify whether the figure contains two panels or an inset for the 1/D extrapolation.","section":"Fig. 4 caption"},{"comment":"The sentence 'we begin our investigations with a 2 × 2 unit cell, allowing four independent tensors at each site to allow the appearance of more general magnetic structures' is awkward; it should read 'allowing four independent tensors, one at each site.'","section":"Section III B"},{"comment":"In the discussion of energetics, the phrase 'although mean-field theory is not a variational approach, even the raw iPEPS energies (without extrapolation) are significantly lower than the mean-field ones' is confusing; the point is that the iPEPS variational upper bound lies below the mean-field energy, which demonstrates the superiority of the iPEPS state. Please rephrase.","section":"Section III C"}],"recommendation":"major_revision","confidential_remarks":"The paper is a single-author numerical study. The author's own previous mean-field work (Refs. [27,47]) is used only for comparison, which is appropriate. The main concern is that the central claim of a diagonal AF stripe ground state is based on a marginal energy difference in a 2×2 supercell, and the larger-supercell evidence is not quantified. The author should be encouraged to provide the missing energy data and a more detailed extrapolation analysis; if the result survives, it would be a nice contribution for a condensed-matter theory journal. Fit with the journal scope is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Worth a look. The genuinely new thing here is a diagonal antiferromagnetic stripe ground state for the 2D periodic Anderson model at n=1.5, which no prior VMC, mean-field, or DMFT study found. The half-filling benchmark against QMC is solid: critical hybridization V~1.2 matches QMC's 1.1(1) and the f magnetization at V=1 matches within error bars. That gives real confidence in the iPEPS implementation. The open data on Zenodo is a plus.\n\nThe soft spot is exactly where you'd expect: the AF vs F energy difference at V=1 is extrapolated to be ~0.015 per site, only about two sigma, and the paper openly says error bars overlap for V<0.8. The larger 2x4 and 4x4 runs are described as exploratory, and no energies are reported. So a vertical stripe, which mean-field actually predicts, or a longer-period diagonal stripe, cannot be ruled out from the reported data. The author does state that the diagonal pattern appeared in all larger-cell runs and that total magnetization extrapolates to zero, which is reassuring, but without energy numbers it's not a complete check. That is the load-bearing gap and the paper should be revised to fill it.\n\nThe paper is otherwise careful: it compares with mean-field, discusses the relation to constrained-path QMC's resonating spin-density-wave state, and is honest about the discrepancy. The extrapolation method (cost function w) is appropriate for iPEPS. The discussion of why mean-field favors vertical stripes while iPEPS does not is reasonable.\n\nWho is this for? People working on heavy-fermion models or on tensor-network methods for two-band systems. It deserves a serious referee. My recommendation: send it to review, but the referee should push for concrete larger-supercell energies and ideally an independent check (DMRG on cylinders) before the central claim is taken as established.","headline":"Solid iPEPS study with a genuinely new diagonal AF stripe at n=1.5, but the ground-state assignment hinges on a small 2x2-supercell energy gap and needs larger-cell energies to close the case.","tokens_in":13823,"tokens_out":2134,"would_cite":true,"duration_ms":19809,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["71.10.Fd","71.27.+a","75.10.-b"],"model":"deepseek-v4-flash","headline":"At 1.5 electrons per site, the two-dimensional periodic Anderson model orders into a diagonal antiferromagnetic stripe, not the uniform ferromagnet predicted by earlier methods.","keywords":["periodic Anderson model","iPEPS","tensor network","magnetic stripe","antiferromagnetism","heavy fermion","square lattice","quantum Monte Carlo benchmark"],"falsifier":"Run iPEPS (or another unbiased method) with $4\\times4$ and larger supercells at $U_f=4$, $V=1$, $\\varepsilon_f=-2$, $n=1.5$, and explicitly compare the extrapolated ($D\\to\\infty$) energies of the diagonal AF stripe, the diagonal F stripe, a vertical stripe, and any incommensurate state; if a non-diagonal state has lower energy, the central claim is refuted.","tokens_in":12905,"feed_emoji":"🧲","tokens_out":11173,"duration_ms":93336,"temperature":0.7,"pith_summary":"This paper claims that the periodic Anderson model on the square lattice, at a filling of $n=1.5$ electrons per site, has a diagonal antiferromagnetic stripe ground state rather than the uniform ferromagnet favored by earlier mean-field and variational studies. The case rests on infinite projected entangled-pair state (iPEPS) calculations, a variational tensor-network method that works directly in the thermodynamic limit and does not suffer the sign problem that limits quantum Monte Carlo away from half-filling. At half-filling the method is benchmarked against quantum Monte Carlo: the antiferromagnet-to-Kondo-singlet transition at $V_{\\rm crit}\\approx 1.2$ and the $f$-electron magnetization match the reference data. At $n=1.5$ the extrapolated energies put the diagonal antiferromagnetic stripe, $E_{\\rm AF} = -3.766(7)$, below the competing diagonal ferromagnetic stripe, $E_{\\rm F} = -3.751(4)$, for $U_f=4$, $V=1$, $\\varepsilon_f=-2$, while for smaller $V$ the two error bars overlap and only the raw variational ordering favors the antiferromagnet.","feed_headline":"Diagonal stripe is the ground state of a 2D heavy-fermion model","feed_subtitle":"Tensor-network simulations, benchmarked against quantum Monte Carlo, find the stripe beats ferromagnet and mean-field orders.","key_machinery":"The central object is the infinite projected entangled-pair state (iPEPS), a tensor network that tiles the infinite square lattice with a small supercell of rank-5 tensors; the two orbitals per site are merged into a 16-dimensional supersite, and the tensors are optimized by imaginary-time evolution with the fast-full-update algorithm and gauge fixing. Accuracy is controlled by the bond dimension $D$ (up to 11 with U(1) spin and charge symmetry), and the infinite environment is contracted with the corner transfer matrix method at $\\chi=D^2$. The decisive step for the ground-state claim is the extrapolation of the variational energy to the $D\\to\\infty$ limit using the normalized cost function $w$ of the time evolution rather than $1/D$, a procedure that makes the small energy difference between the diagonal AF and F stripes resolvable.","core_discovery":"At $n=1.5$ filling with $U_f=4$, $V=1$, and $\\varepsilon_f=-2$, the paper identifies the ground state as a diagonal antiferromagnetic stripe of the localized $f$-electron moments: the moments order antiferromagnetically along the $[1\\bar{1}]$ direction, while along the $[11]$ direction the magnetization is almost zero ($|m_f^i|\\sim 0.09$ versus $0.4$) and the local correlation $C=|\\langle \\hat n_{j\\uparrow}^f \\hat n_{j\\downarrow}^f\\rangle - \\langle \\hat n_{j\\uparrow}^f\\rangle\\langle \\hat n_{j\\downarrow}^f\\rangle|$ is strongly enhanced, signaling that those sites stay nonmagnetic through local charge-spin fluctuations rather than through a uniform paramagnet. Extrapolating iPEPS energies to the error-free limit with a third-order polynomial in the cost function gives $E_{\\rm AF}=-3.766(7)$ for this state and $E_{\\rm F}=-3.751(4)$ for the symmetry-related diagonal ferromagnetic stripe, so the AF stripe is the ground state at this coupling; at $V\\lesssim 0.8$ the extrapolated error bars overlap and the AF state is preferred only at the variational level. The same diagonal stripe pattern reappears in exploratory $2\\times4$ and $4\\times4$ supercells, and the total supercell spin flows to zero as the bond dimension grows.","pith_inferences":["Beyond the paper, the resonating spin-density-wave ground state seen in constrained-path quantum Monte Carlo on small clusters could be the finite-size precursor of this diagonal AF stripe; a large-scale unbiased simulation that can resolve long-range order would settle whether the two are the same phase.","Beyond the paper, the diagonal-versus-vertical stripe competition mirrors the stripe physics of the two-dimensional Hubbard model near 1/8 doping, except that here the two competing stripe polarities (AF versus F) are being selected rather than two stripe orientations.","Beyond the paper, a testable extension is to dope away from $n=1.5$ and check where the diagonal stripe gives way to a paramagnetic Kondo state; the paper's observation that magnetism dies near $n_f\\sim0.5$ suggests the same depletion mechanism seen at half-filling.","Beyond the paper, adding Hund's coupling through a Kanamori interaction could change the energy balance between the AF and F stripes, since Hund's coupling directly favors ferromagnetic alignment of the local moments; the paper lists this as an outlook."],"forward_implications":["If the diagonal AF stripe is the true ground state at $n=1.5$, earlier mean-field and variational Monte Carlo predictions of a uniform ferromagnet at this filling miss the actual ordering and need larger unit cells or more general trial states.","The near-degeneracy of the diagonal ferromagnetic stripe means small changes in hybridization or $f$-level energy could switch the ordering, so the model likely hosts an AF-to-F stripe transition inside the $n=1.5$ sector.","The half-filling benchmark against quantum Monte Carlo establishes that iPEPS produces quantitatively reliable critical points and magnetizations for this two-band model, backing its use in sign-problem-affected parameter regions.","Because the conduction electrons remain almost unpolarized ($m_c \\sim 10^{-3}$) while the $f$-electron moments order, the magnetic stripe is carried by the localized band with the conduction band responding only through correlations.","The absence of vertical stripes in iPEPS, despite mean-field favoring them, indicates that local suppression of double occupancy, not just magnetic order, is essential to the ground-state energy at this filling."],"supporting_citations":[{"why":"Supplies the quantum Monte Carlo benchmark for half-filling; the paper compares its critical hybridization and f-electron magnetization to this reference.","marker":"[12]"},{"why":"Provides the constrained-path quantum Monte Carlo results at n=1.5 whose resonating spin-density-wave state is discussed as a possible precursor of the diagonal AF stripe.","marker":"[22]"},{"why":"Represents the variational Monte Carlo prediction of a ferromagnetic ground state at n=1.5 that the paper's diagonal stripe result challenges.","marker":"[14]"},{"why":"Define the iPEPS tensor-network framework used for all simulations, including the supercell ansatz and contraction scheme.","marker":"[29-34]"},{"why":"Introduces the fast-full-update algorithm with gauge fixing, which allows bond dimensions up to D=11 and defines the cost function used for extrapolation.","marker":"[43]"},{"why":"Advocates energy extrapolation as a function of the normalized cost function, the procedure the paper uses to rank the AF and F stripe states.","marker":"[46]"}],"fun_headline_variants":["Diagonal stripe wins as ground state in 2D heavy-fermion model","Novel diagonal antiferromagnetic stripe phase found in 2D Anderson model","At n=1.5, diagonal stripe beats ferromagnet in Anderson lattice","iPEPS simulation reveals stripe ground state in Anderson model","Antiferromagnetic diagonal stripe is ground state at V=1 in 2D model"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that a four-site repeating cell, with only exploratory checks of eight- and sixteen-site cells, contains the true magnetic pattern at $n=1.5$; if the true order repeats over a longer distance (for example as a vertical stripe), the diagonal-stripe conclusion collapses.","fun_headline_variants_meta":{"raw":{"variants":["Diagonal stripe wins as ground state in 2D heavy-fermion model","Novel diagonal antiferromagnetic stripe phase found in 2D Anderson model","At n=1.5, diagonal stripe beats ferromagnet in Anderson lattice","iPEPS simulation reveals stripe ground state in Anderson model","Antiferromagnetic diagonal stripe is ground state at V=1 in 2D model"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00034,"raw_usage":{"total_tokens":1897,"prompt_tokens":986,"completion_tokens":911,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":602,"completion_tokens_details":{"reasoning_tokens":811}},"tokens_in":602,"tokens_out":911,"duration_ms":8859,"temperature":1.0,"reasoning_tokens":811,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:38:25.577159+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run iPEPS (or another unbiased method) with $4\\times4$ and larger supercells at $U_f=4$, $V=1$, $\\varepsilon_f=-2$, $n=1.5$, and explicitly compare the extrapolated ($D\\to\\infty$) energies of the diagonal AF stripe, the diagonal F stripe, a vertical stripe, and any incommensurate state; if a non-diagonal state has lower energy, the central claim is refuted.","supporting_citations":[{"cited_title":"Veki´ c, J","cited_arxiv_id":null,"evidence_quote":"Supplies the quantum Monte Carlo benchmark for half-filling; the paper compares its critical hybridization and f-electron magnetization to this reference."},{"cited_title":"Kubo, Lifshitz transitions in magnetic phases of the periodic anderson model, Journal of the Physical Society of Japan 84, 094702 (2015)","cited_arxiv_id":null,"evidence_quote":"Represents the variational Monte Carlo prediction of a ferromagnetic ground state at n=1.5 that the paper's diagonal stripe result challenges."}],"review_version":1}