{"id":"edd36a0f-0f2a-4dad-9b21-84b071e5ff07","arxiv_id":"1908.07502","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"A six-loop field-theoretic calculation gives the fractal dimension of critical curves and the crossover exponent in O(n) models, with estimates for LERW, SAW, Ising, XY and Heisenberg systems.","lead":"This paper computes the fractal dimension of critical curves in O(n) symmetric field theory to six-loop order, covering loop-erased random walks, self-avoiding walks, and Ising and XY lines. It also introduces a self-consistent resummation method and uses exact two-dimensional results to improve three-dimensional predictions for critical exponents.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Claimed backbone fractal dimension exceeds the dimension of the all-lines set in which it must be contained, for n>0.","rationale":"The reader identified the operator-to-geometry correspondence (Eqs. (35)-(37)) as the weakest premise, and I agree that is where the argument is most exposed. However, the reader did not notice that the paper's own numbers contradict that correspondence: the claimed backbone dimension d_f exceeds the claimed all-lines dimension d_tot^f=1/ν for n=1 and n=2, even though the backbone is defined as a subset of all lines. This is not a failure of resummation precision or an outside-consensus disagreement; it is an internal inconsistency between Eqs. (34), (37), and (40), and it is large: 1.735 versus 1.588 in d=3 for the Ising case. The only ways out are either that the operator tilde E does not measure the geometric length of the backbone (e.g., it measures a signed quantity whose absolute scaling differs), or that d_tot^f=1/ν is not the fractal dimension of the union of lines for n>0. Both options undermine the central claim that the computed d_f values are the fractal dimensions of critical curves. The agreement with simulations for Ising and XY then becomes coincidental or the simulations measure a different geometric object. The paper still contains a valid six-loop calculation of a crossover exponent, but the headline geometric interpretation and the title claim are not supported. Hence the reader's ACCEPT should be revised downward; the central claim is sufficiently compromised that REJECT is the appropriate verdict for the paper as stated, with the possibility of a corrected version focusing on the crossover exponent alone.","tokens_in":22548,"tokens_out":41034,"duration_ms":500987,"concrete_test":"Extract from the published data of Winter, Janke and Schakel (Phys. Rev. E 77, 061108) the measured fractal dimension of the full high-temperature graph (propagator line plus all attached loops) for the 3D Ising model. If this dimension is at least 1.735, then the paper's d_tot^f=1/ν≈1.587 is not the dimension of the union of all lines, invalidating Eq. (34) as a geometric statement. If instead the union dimension is about 1.587, then the propagator line, being a subset, cannot have dimension 1.735, invalidating Eq. (37) as the backbone dimension. Additionally, re-evaluate the four-loop Kirkham series for γ_tildeE and γ_1 at n=1 to check the sign of γ_tildeE(g*)-γ_1(g*) in d=3; the geometric bound requires this difference to be non-positive.","verdict_should_be":"REJECT","load_bearing_attack":"The central identification is Eqs. (34) and (37): d_tot^f = 2+γ_1(g*)-η = 1/ν is the fractal dimension of all lines (backbone plus surrounding loops), and d_f = 2+γ_tildeE(g*)-η is the fractal dimension of the backbone. Since the backbone is a subset of the all-lines set, any valid geometric interpretation requires d_f ≤ d_tot^f. The paper's own d=3 numbers violate this: for n=1, Fig. 2 gives d_f=1.7353(10) while Table VI gives ν=0.6296(3), so 1/ν≈1.588; for n=2, d_f=1.7644(10) versus 1/ν≈1.49. The same violation occurs in d=2, where the cited CFT values give d_f(Ising)=11/8=1.375 while 1/ν=1. Equation (40) even states 'we expect d_tot^f > d_f' and simultaneously 'φ'_c(n)>0'; these are incompatible, since φ_c=d_f/d_tot^f. The numerical results choose φ'_c>0 and thus d_f>d_tot^f. This is not a resummation artifact: the fixed-point RG functions themselves give γ_tildeE(g*) > γ_1(g*) for n>0, making the traceless-operator dimension exceed the trace-operator dimension. Therefore either tilde E measures a signed length whose scaling is not the geometric backbone length, or d_tot^f=1/ν is not the dimension of the union of lines. In both cases, the operator-to-geometry correspondence underlying the headline d_f values for Ising and XY is unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper computes the six-loop renormalization-group function γ_\\tilde{E} for the traceless bilinear operator in the O(n)-symmetric φ^4 theory in d = 4−ε (Eq. (41)), proposes d_f = 2 + γ_\\tilde{E}(g*) − η as the fractal dimension of the 'propagator' or 'backbone' line of critical curves (Eq. (37)), and evaluates d_f for n = −2, 0, 1, 2 in d = 3, claiming agreement with simulations of loop-erased random walks, self-avoiding walks, and Ising and XY propagator lines. The same operator is used to define the crossover exponent φ_c = ν d_f = d_f / d_tot^f with d_tot^f = 1/ν (Eqs. (39), (54)-(55)). A new self-consistent Borel-type resummation (SC) is introduced and combined with the KP17 resummation. The results are cross-checked against the large-n expansion (Section VIII) and against exact d=2 conformal field theory results (Section VI), yielding improved d=3 estimates for d_f, ν, η, ω and φ_c.","tokens_in":22893,"tokens_out":39305,"duration_ms":385739,"significance":"If correct, the paper provides a unified six-loop description of the fractal geometry of O(n) critical lines, and the numerical agreement with independent high-precision simulations is genuinely impressive: LERW 1.6243(10) vs 1.62400(5), SAW 1.7027(10) vs 1.701847(2), Ising 1.7353(10) vs 1.7349(65), XY 1.7644(10) vs 1.7655(20). The verification of the ε-expansion of φ_c against the known O(1/n²) large-n result (Section VIII) is a strong and non-trivial consistency check, as is the stated agreement with Kirkham's four-loop result. The d=2 comparison (5/4, 4/3, 11/8, 3/2) shows that the ε-expansion, pushed to ε=2, captures the correct qualitative trend. The SC scheme is heuristic, but it is cross-validated by the independent KP17 method and by the explicit α-bounds in Fig. 7, and the authors are honest about its limitations. The paper makes falsifiable predictions for d_f and φ_c in d=3 that are directly comparable to simulation and experiment. These strengths make the paper worth publishing, provided the interpretive inconsistency identified below is resolved.","major_comments":[{"comment":"The stress-test concern that the claimed backbone dimension exceeds the all-lines dimension lands, and it is verifiable from the paper's own numbers. Eq. (34) identifies d_tot^f = 1/ν with the fractal dimension of 'all lines' (backbone plus loops), Eq. (37) gives the backbone dimension d_f = 2 + γ_\\tilde{E}(g*) − η, and the text before Eq. (40) asserts that the backbone is contained in the union of backbone plus loops. Monotonicity of the fractal dimension then requires d_f ≤ d_tot^f. The reported values violate this for every n>0: for n=1, d_f = 1.7353(10) (Fig. 2) while 1/ν = 1.5883(8) (Table VI); for n=2, d_f = 1.7644(10) while 1/ν = 1.4912(5); equivalently φ_c = νd_f > 1 (Table IV: 1.089(1), 1.180(4), 1.265(5) for n=1,2,3), which is d_f/d_tot^f > 1. This is not a resummation artifact: from the leading term of Eq. (41), γ_\\tilde{E} = −2ε/(n+8) + O(ε²), and the standard ν^{-1} = 2 − (n+2)ε/(n+8) + O(ε²), one obtains d_f − d_tot^f = γ_\\tilde{E}(g*) − γ_1(g*) = nε/(n+8) + O(ε²) > 0 for n>0 already at one loop. The manuscript is also internally contradictory: Eq. (40) asserts both 'd_tot^f > d_f' and 'φ'_c(n) > 0' with φ_c(0)=1, but since φ_c = d_f/d_tot^f, the first statement implies φ_c < 1 and hence φ'_c(0) < 0. The exact d=2 results are likewise inconsistent with the subset interpretation (Ising: d_f = 11/8 > 1/ν = 1; XY: d_f = 3/2 while ν diverges, Fig. 11). The authors must state which identification, (34) or (37), fails for n>0, explain what geometric object (if any) has dimension 1/ν for n>0, and correct Eq. (40). Until then, the interpretation of d_f for the Ising and XY cases as a geometric fractal dimension of the backbone is unsupported, even though the d_f values themselves agree impressively with the simulations.","section":"§I, Eqs. (34), (37), (39)-(40), Fig. 2, Table VI"},{"comment":"The six-loop function (41) is the central new technical object of the paper, but it is presented without a derivation: no diagram-by-diagram count, no integration method, and no supplementary material is provided for the coefficients involving ζ3,5, ζ5, ζ7 and ζ9. The stated agreement with Kirkham's four-loop result and with the O(1/n²) large-n expansion (Section VIII) checks only low orders and a partial large-n sector, so the ε⁵ and ε⁶ terms at fixed n, which are the genuinely new content, cannot currently be verified from the manuscript. I request that the authors supply the computation in reproducible form (a diagram list with symmetry factors, or a code and data supplement), or at minimum a documented derivation of the non-trivial constants such as the ζ3,5 and ζ3² coefficients.","section":"§II, Eq. (41) and §III"},{"comment":"The headline d=3 error bars (for example d_f(SAW) = 1.7027(10), d_f(Ising) = 1.7353(10), d_f(XY) = 1.7644(10) in Fig. 2) are produced by the new SC scheme, whose status is explicitly qualified in §III: the text says the error bars 'have to be taken with a grain of salt', and for the LERW case the α-scan yields only a range d_f ∈ [1.62378, 1.6254] from which the central value 1.62426 is taken as the mean (Fig. 7). The paper should specify exactly how the reported uncertainties are derived from the α-bounds (mean, midpoint, or spread), and should discuss whether the d=3 error bars could be underestimated in the same way that the d=2 values are, where different resummation schemes scatter by about 0.05 (Fig. 3). The agreement with simulation is robust enough that this does not change the main conclusions, but the quoted precision currently overstates the rigor of the method.","section":"§III, Figs. 2 and 7"}],"minor_comments":[{"comment":"The abstract ('in agreement with numerical simulations') and the statement in §I that the agreement 'firmly establishes that the appropriate operator was identified' are too strong while the d_f > d_tot^f issue of Major 1 is unresolved; please qualify these claims.","section":"Abstract and §I"},{"comment":"The symbol \\tilde{E} denotes three different objects: the traceless tensor of Eq. (29), the integrated φ_1²−φ_2² insertion of Eq. (35), and the anisotropic mass combination of Eq. (51). The authors acknowledge this, but distinct notations would substantially improve readability.","section":"§I, Eq. (35) and §IV, Eq. (51)"},{"comment":"References [1], [2], [13] and [22] contain garbled strings ('V olume', 'exponants'); please correct the bibliography entries.","section":"References"},{"comment":"The table in Fig. 3 quotes single-scheme values with single-scheme errors (for example 1.416(1) for Ising in d=2), while the caption states that the overall error is of order 0.05; please make the displayed errors reflect the global estimate, or mark the values clearly as scheme-specific.","section":"Fig. 3"},{"comment":"The discussion of ω is acknowledged to be inconclusive ('It is not even clear whether this is a question which can be answered via CFT'), yet Table VII reports ω values with errors as small as 0.004; a sentence clarifying that ω is not part of the paper's central claims would help prevent overinterpretation.","section":"§VI A and Fig. 13"}],"recommendation":"major_revision","confidential_remarks":"I verified the O(ε) statement in Major 1 independently: with γ_\\tilde{E} = −2ε/(n+8) and ν^{-1} = 2 − (n+2)ε/(n+8), the difference is nε/(n+8), so the d_f > d_tot^f violation is built into the one-loop theory and is not a resummation artefact; the contradiction with the manuscript's own Eq. (40) is textual. The numerical values are nevertheless very likely correct given the agreement with several independent simulations and with the large-n expansion; the issue is in the interpretation, which the authors can fix with a careful revision. Please also press for the reproducibility of Eq. (41). The reliance on the authors' own refs. [43,44] for the n = −2 mapping is acceptable here because the resulting d_f matches Wilson's independent simulation to about 0.02%."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: the six-loop RG calculation is real and worth refereeing, but the geometric interpretation for n>0 is not supported. The numbers violate the basic subset inequality d_f <= d_tot, and Eq. (40) states expectations that contradict each other.\n\nWhat is actually new: Eq. (41) extends Kirkham's four-loop gamma_tildeE to six loops, Eq. (56) gives the crossover exponent series, and the large-n check is a genuine cross-validation. The self-consistent resummation is inventive and gives excellent results for loop-erased random walks (n=-2) and self-avoiding walks (n=0), where the operator mapping is on safe ground.\n\nWhere it falls apart: for Ising (n=1) and XY (n=2) the reported d_f values in d=3 (1.7353 and 1.7644) are larger than the all-lines dimension 1/nu (1.588 and 1.491). A subset cannot have a larger Hausdorff dimension than its superset. The paper even writes 'we expect d_tot^f > d_f' and then immediately 'phi'_c(n)>0'; since phi_c=d_f/d_tot, these cannot both hold. That is not a resummation artifact; it is a contradiction in the operator interpretation. Either tilde E measures a signed length whose scaling is not the geometric backbone length, or 1/nu is not the dimension of the union of lines. In either case the central claim for n>0 is unsupported. The same issue appears in d=2, where their own CFT values give d_f=11/8 > 1/nu=1 for Ising.\n\nMinor: the six-loop series is stated without a diagram-by-diagram derivation, no code or data files are shipped, and the SC error bars are explicitly heuristic. These are minor compared to the interpretation problem.\n\nWho this is for: people working on O(n) RG functions and crossover exponents. The phi_c series and the large-n expansion are likely correct and useful. But do not cite the Ising/XY d_f values as geometric facts until the identification is fixed.\n\nRecommendation: send it to peer review. The RG computation is substantial and deserves referee time, but the referee should demand the authors address the subset inequality and clarify what tilde E actually measures.","headline":"The six-loop RG calculation is real and worth refereeing, but the geometric interpretation for n>0 is not supported: the reported backbone dimension exceeds the dimension of the all-lines set, and Eq. (40) states mutually incompatible expectations.","tokens_in":23440,"tokens_out":8011,"would_cite":false,"duration_ms":78838,"reading_group":"maybe","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["05.10.Cc","64.60.Fr","11.10.Gh"],"model":"deepseek-v4-flash","headline":"A traceless operator insertion gives the fractal dimension of critical curves in the O(n) $\\phi^4$ model, matching LERW, SAW, Ising and XY simulations.","keywords":["fractal dimension","critical curves","O(n) phi^4 model","loop-erased random walks","self-avoiding walks","crossover exponent","epsilon expansion","renormalization group"],"falsifier":"A decisive test would be a Monte Carlo measurement of the fractal dimension of the propagator lines of the 3D XY model with error below $10^{-3}$: the paper predicts $d_f=1.7644(10)$, and a statistically significant deviation would falsify the operator-to-geometry identification. A second falsifier is a seven-loop computation of $\\gamma_{\\tilde E}$: if the resummed value moves away from the six-loop result instead of stabilizing, the series or the correspondence is wrong.","tokens_in":22275,"feed_emoji":"🌀","tokens_out":15227,"duration_ms":136107,"temperature":0.7,"pith_summary":"The paper sets out to show that the fractal dimension of the curves seen at criticality in the O(n)-symmetric $\\phi^4$ model is not a separate geometric input but is encoded in a single composite operator. Inserting the traceless bilinear operator $\\tilde E$ into a propagator counts the backbone line and drops the surrounding loops, so its anomalous dimension fixes $d_f = 2 + \\gamma_{\\tilde E}(g_*) - \\eta$. The paper computes $\\gamma_{\\tilde E}(g_*)$ to six loops in $d=4-\\epsilon$; at $n=-2,0,1,2$ the resulting numbers for loop-erased random walks, self-avoiding walks, Ising lines, and XY lines agree with numerical simulations in $d=3$. Because the total fractal dimension of all lines is $1/\\nu$, the ratio $d_f/(1/\\nu)=\\nu d_f$ is exactly the crossover exponent of an anisotropic mass term, so the same series predicts $\\phi_c$ for experiments and Monte Carlo. A self-consistent resummation introduced in the paper, guided by exact $d=2$ results, sharpens the $d=3$ estimates for $\\nu$, $\\eta$, $\\omega$, and $\\phi_c$.","feed_headline":"Six-loop field theory gives fractal dimensions of critical curves","feed_subtitle":"One operator insertion reproduces LERW, SAW, Ising, and XY curve dimensions in 3D.","key_machinery":"The load-bearing object is the renormalized traceless bilinear operator $\\tilde E_{ij}=\\phi_i\\phi_j-\\delta_{ij}E$, equivalently its integrated difference form $\\tilde E=\\frac12\\int_y(\\phi_1^2-\\phi_2^2)$, which acts as a geometric filter: inserted on a propagator line it reproduces the mass-operator insertion, while inserted on a loop it gives zero. Its renormalization factor $Z_{\\tilde E}$ defines $\\gamma_{\\tilde E}=\\beta\\partial_g\\ln Z_{\\tilde E}$, and the identity $d_f=2+\\gamma_{\\tilde E}(g_*)-\\eta$ carries the whole argument. The quantitative engine is the six-loop expansion of $\\gamma_{\\tilde E}$ and of the crossover exponent $\\phi_c$, together with a newly proposed self-consistent resummation that fits the asymptotic ratios $b_n/b_{n-1}$ to $a+be^{-cn}$; combining that with the location of $d=2$ singularities fixes the best variables for extrapolation to $d=3$.","core_discovery":"The central claim is a precise operator-to-geometry dictionary. In the renormalized O(n) $\\phi^4$ theory, insert the integrated traceless tensor $\\tilde E=\\frac12\\int_y(\\phi_1^2-\\phi_2^2)$ into a propagator line with a fixed component index: this weights the backbone line exactly as the mass operator does, while on a closed loop, where all indices are summed, it vanishes. The anomalous dimension of $\\tilde E$ therefore measures the fractal dimension of the backbone through $d_f=2+\\gamma_{\\tilde E}(g_*)-\\eta$, and the total set of lines, backbone plus loops, has dimension $1/\\nu$, so $\\phi_c=\\nu d_f$ is both a ratio of fractal dimensions and the crossover exponent of a mass anisotropy. The paper evaluates $\\gamma_{\\tilde E}(g_*)$ explicitly to six-loop order in Eq. (41) and $\\phi_c$ to the same order in Eq. (56). In $d=3$, the predictions are $d_f=1.6243(10)$ for loop-erased random walks ($n=-2$), $1.7027(10)$ for self-avoiding walks ($n=0$), $1.7353(10)$ for Ising lines ($n=1$), and $1.7644(10)$ for XY lines ($n=2$), each consistent with high-precision simulations. In $d=2$, independent resummations bracket the exact CFT values such as $5/4$ for LERW, and the paper uses those exact values to choose resummation variables that improve the $d=3$ estimates.","pith_inferences":["The same operator filter could be applied to other nonlocal geometric observables, such as loop-length distributions, intersection counts, or the size of erased loops, by inserting different composite operators into the same six-loop diagrams; the paper does not carry this out.","Because the paper identifies $\\phi_c$ with $\\nu d_f$, every experimental or Monte Carlo measurement of the crossover exponent in an anisotropic magnet is implicitly a measurement of the backbone fractal dimension; using the published experimental values collected in the paper would give independent checks of $d_f$ for $n=2,3$.","The $\\epsilon$-expansion extrapolations to $d=2$ scatter by about 0.05 around the exact CFT values; a resummation that builds in the square-root singularity at $n=\\pm2$ in the $(d,n)$ plane should collapse that scatter, and the paper lists this as a direction for future work."],"forward_implications":["Loop-erased random walks in three dimensions acquire a field-theoretic prediction, $d_f=1.6243(10)$, matching the precise numerical value $1.62400(5)$, so the non-Markovian loop-erasing process is captured by a local $\\phi^4$ theory at $n=-2$.","For self-avoiding walks, backbone and total lines coincide, so $d_f=1/\\nu$; the paper's $\\nu=0.5874(2)$ is within about $3\\times10^{-4}$ of the best simulations.","The crossover exponent comes out as $\\phi_c=1.089(1)$ for Ising ($n=1$) and $1.180(4)$ for XY ($n=2$) in $d=3$, in line with experiments on anisotropic magnets and structural phase transitions.","Resumming $1/\\nu^3$ and $1/\\phi_c^{13/4}$, the variables suggested by $d=2$ singularities, yields sharper $d=3$ estimates for $\\eta$ and $\\omega$ as well as for $d_f$."],"supporting_citations":[{"why":"Supplies the minimal-subtraction six-loop renormalization functions of the O(n)-symmetric $\\phi^4$ theory, and its resummation method is the baseline against which the new self-consistent scheme is compared.","marker":"[10]"},{"why":"Defines the crossover exponent through the anisotropic-mass O(k) times O(n-k) action and gives the scaling form whose ratio of dimensions is $\\phi_c=\\nu d_f$.","marker":"[11]"},{"why":"Provides the exact conformal-invariance result $d_f=5/4$ for planar loop-erased random walks, the $d=2$ anchor used to test the $\\epsilon$-expansion.","marker":"[41]"},{"why":"Establishes the equivalence between the O(n) $\\phi^4$ theory at $n=-2$ and loop-erased random walks in all dimensions, which justifies reading $d_f$ at $n=-2$ as the LERW dimension.","marker":"[43]"},{"why":"Provides the high-precision numerical value $d_f=1.62400(5)$ for the three-dimensional loop-erased random walk, the benchmark for the $n=-2$ prediction.","marker":"[45]"},{"why":"Gives the earlier four-loop expansion of the crossover exponent that the six-loop series extends and corrects at order $\\epsilon^3$.","marker":"[48]"},{"why":"Supplies high-precision Monte Carlo data for self-avoiding walks in three dimensions, used to benchmark the $n=0$ values of $\\nu$ and $d_f$.","marker":"[24]"},{"why":"Provides simulations of the geometric properties of lines in the three-dimensional Ising and XY models, the benchmark for $n=1$ and $n=2$.","marker":"[46]"},{"why":"Gives the large-n expansion of the crossover exponent to $1/n^2$; expanding both it and the $\\epsilon$-series in a double expansion provides a strong consistency check.","marker":"[49]"}],"fun_headline_variants":["Inserting one operator yields LERW, SAW, Ising, and XY fractal dimensions","Crossover exponent from a single operator insertion in O(n) phi^4","Self-consistent resummation sharpens six-loop fractal dimensions","Operator geometry: one trace insertion sets all fractal curve dimensions","Six-loop epsilon expansion fixes fractal dimension of LERW, SAW, Ising, XY"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire identification of $d_f$ with a field-theory anomalous dimension rests on the assumption that the traceless operator insertion marks exactly the backbone curve and ignores the loops; if that geometric reading fails in three dimensions, the six-loop agreement with simulations would be coincidental.","fun_headline_variants_meta":{"raw":{"variants":["Inserting one operator yields LERW, SAW, Ising, and XY fractal dimensions","Crossover exponent from a single operator insertion in O(n) phi^4","Self-consistent resummation sharpens six-loop fractal dimensions","Operator geometry: one trace insertion sets all fractal curve dimensions","Six-loop epsilon expansion fixes fractal dimension of LERW, SAW, Ising, XY"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000962,"raw_usage":{"total_tokens":4191,"prompt_tokens":1136,"completion_tokens":3055,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":752,"completion_tokens_details":{"reasoning_tokens":2953}},"tokens_in":752,"tokens_out":3055,"duration_ms":21842,"temperature":1.0,"reasoning_tokens":2953,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:14:44.387978+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive test would be a Monte Carlo measurement of the fractal dimension of the propagator lines of the 3D XY model with error below $10^{-3}$: the paper predicts $d_f=1.7644(10)$, and a statistically significant deviation would falsify the operator-to-geometry identification. A second falsifier is a seven-loop computation of $\\gamma_{\\tilde E}$: if the resummed value moves away from the six-loop result instead of stabilizing, the series or the correspondence is wrong.","supporting_citations":[{"cited_title":"Falconer, The Geometry of Fractal Sets, Cambridge Uni- versity Press, Cambridge, U.K., 1986","cited_arxiv_id":null,"evidence_quote":"Provides the exact conformal-invariance result $d_f=5/4$ for planar loop-erased random walks, the $d=2$ anchor used to test the $\\epsilon$-expansion."},{"cited_title":"De Gennes, Exponents for the excluded volume problem as derived by the Wilson method, Phys","cited_arxiv_id":null,"evidence_quote":"Establishes the equivalence between the O(n) $\\phi^4$ theory at $n=-2$ and loop-erased random walks in all dimensions, which justifies reading $d_f$ at $n=-2$ as the LERW dimension."},{"cited_title":"Hausdorff di- mension of critical fluctuations in Abelian gauge theories","cited_arxiv_id":null,"evidence_quote":"Gives the earlier four-loop expansion of the crossover exponent that the six-loop series extends and corrects at order $\\epsilon^3$."},{"cited_title":"Scale-free Monte Carlo method for calculating the critical exponent $\\gamma$ of self-avoiding walks","cited_arxiv_id":"1701.08415","evidence_quote":"Supplies high-precision Monte Carlo data for self-avoiding walks in three dimensions, used to benchmark the $n=0$ values of $\\nu$ and $d_f$."},{"cited_title":"Kirkham, Calculation of crossover exponent from Heisen- berg to Ising behaviour using the fourth-order ϵ expansion, J","cited_arxiv_id":null,"evidence_quote":"Gives the large-n expansion of the crossover exponent to $1/n^2$; expanding both it and the $\\epsilon$-series in a double expansion provides a strong consistency check."}],"review_version":1}