{"id":"a81a2dc7-425b-4481-9ebb-128e8caa5a49","arxiv_id":"2412.00931","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":11,"one_line_summary":"A Horndeski scalar-field dark energy model that crosses the phantom divide and has negative early-time energy density fits cosmological data as well as LambdaCDM but does not resolve the main tensions.","lead":"The paper tests a modified-gravity dark energy model in which a scalar field crosses the phantom divide and has negative energy density at early times, while remaining stable. It constrains the model with CMB, supernova, and BAO data and finds it fits as well as the standard cosmological model, but without resolving the Hubble or S8 tensions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reported 1σ preference for nonzero c1,c2,c3 is conditional on the unvaried initial field value φ_i=10 and linear potential; varying φ_i rescales the physical effect of c3 and may erase the preference.","rationale":"I read the paper as a proof-of-concept for stable Horndeski dark energy with phantom crossing and negative ρφ; that part is supported by explicit background and stability expressions and by hi_class outputs. The fit is honestly not better than ΛCDM (Δχ²=-0.6 for three extra parameters), so the lasting claim is the qualitative features plus the weak positive-coupling preference. The most insecure condition for that claim is the fixed initial value φ_i=10. In scalar-tensor theories a constant shift of φ is physical because G4=1/2+c3φ and V=V0φ; there is no shift symmetry. Thus the data preference for c3 is degenerate with the choice of φ_i. If φ_i were varied, the posterior distributions for c1, c2, c3 could change materially, and the conclusion that this class of models merits further exploration would be weakened. The reader's weakest-assumption identification is essentially correct, and I sharpen it by noting the precise rescaling mechanism: the physical effect of c3 is set by the product c3φ_i. I do not see an internal inconsistency in the stability algebra or the perturbation equations; the absence of released code is a secondary issue. Therefore the verdict stays CONDITIONAL, not because the theory is flawed but because the data-preference claim is not yet shown robust to a physically meaningful model choice.","tokens_in":20766,"tokens_out":10267,"duration_ms":101420,"concrete_test":"Rerun the MCMC with identical likelihoods and priors but with φ_i ∈ {1, 10, 100} and φ'_i ∈ {10^-12, 10^-10, 10^-8} at z≈1000, re-tuning V0 by shooting in each case, and also with V(φ)=V0φ^2 for one φ_i. If the 1σ lower bounds on c1, c2, c3 in Table I do not remain positive for all configurations, the claimed data preference is an artifact of the chosen initial condition/potential and should be reported as model-setup-dependent, not as a property of the Horndeski class. A minimal version: compute the best-fit χ² and c3 posterior with φ_i=1 versus φ_i=10; a shift in c3 by the expected factor ~10, or a change in sign of the lower bound, settles the concern.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central observational claim—that 'there remains a preference for non-zero values of the model parameters'—rests on a single unvaried choice for the scalar field initial conditions at z≈1000: φ_i=10, φ'_i=10^-10 (Sec. III), together with the linear potential V=V0φ. This is not a harmless convention. The nonminimal coupling enters the background as G4=1/2+c3φ, so the physical strength of c3 is set relative to the chosen φ_i. At the initial epoch, the best-fit c3≈4×10^-4 gives δG4/G4≈2c3φ_i≈0.008; for φ_i=1 the same physical G4 would require c3≈4×10^-3, and the MCMC posterior on c3 would shift accordingly. Because φ_i is neither a symmetry nor an attractor value and is not sampled, the reported 1σ positive preference for c3 (and its connection to negative high-redshift energy density) is conditional on this ansatz. A non-linear potential would further alter the late-time dynamics and the c1/c2 constraints. The theoretical ability of the Lagrangian to exhibit phantom crossing is not in question; what is load-bearing is the data-driven preference, which cannot be separated from the unvaried initial condition.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a Horndeski dark energy model defined by G2 = X − V0φ, G3 = c1φ + c2X, G4 = 1/2 + c3φ, and G5 = 0. It demonstrates that this model can produce phantom crossing and negative scalar-field energy density at high redshifts while maintaining Qs > 0 and cs^2 > 0, and it studies the impact on the growth rate, matter power spectrum, and CMB temperature spectrum. The authors perform an MCMC analysis with Planck18, BAO/fσ8, and PantheonPlus data, finding a fit comparable to ΛCDM (Δχ² = −0.6) and reporting a 1σ preference for positive nonzero c1, c2, and c3.","tokens_in":21018,"tokens_out":9286,"duration_ms":84285,"significance":"If the results are sound, the paper provides a concrete, non-parametric dark energy construction in which phantom crossing and negative energy density arise from a fundamental scalar-tensor action rather than from a phenomenological parameterization. The use of the public hi_class and MontePython codes and of standard likelihoods is a strength, and the stability conditions are stated explicitly. The quantitative claim of an observational preference for nonzero couplings is, however, weak and conditional on several unvaried choices, so the paper's main value is as a proof-of-concept and a starting point for more robust tests. The comparison with ΛCDM is honest, and the analysis correctly emphasizes that perturbation data, especially CMB, strongly constrain the model.","major_comments":[{"comment":"The sign convention for X is inconsistent. Section II defines X = ∂µϕ∂µϕ/2, and Appendix A uses the FLRW metric ds² = −dt² + a²dx², for which ∂µϕ∂µϕ = −φ̇², so X = −φ̇²/2. With this convention, the kinetic term in G2 = X − V is negative, contradicting the claim that the c1 = c2 = c3 = 0 limit is canonical quintessence. The sign inconsistency propagates to Eq. (5) and to the stability conditions (A20)–(A21), where the signs of the c1 and c2 terms depend on this convention. Since the central claims about c2 ensuring Qs > 0 and about the sign of the c3 contribution to ρϕ rest on these signs, the authors must adopt a single, explicit convention (e.g., X = −(1/2)∂µϕ∂µϕ) and re-derive Eqs. (5), (A20), and (A21) accordingly.","section":"Sec. II and Appendix A"},{"comment":"The reported 1σ preference for nonzero c1, c2, and c3 is conditional on the fixed initial conditions φi = 10 and φ̇i = 10⁻¹⁰ set at z ≈ 1000 and on the linear potential V = V0φ. The physical strength of the nonminimal coupling is set by the combination c3φ, so scaling φi by an order of magnitude rescales the effective coupling; the posterior on c3 would shift correspondingly. The paper does not vary φi, φ̇i, or the potential shape, and no robustness tests are presented. Without such tests, the data-driven preference for these couplings cannot be regarded as a robust model prediction, even though the theoretical capability of the model to exhibit phantom crossing is not in question.","section":"Sec. III and Table I"},{"comment":"The MCMC convergence criterion R − 1 ≲ 0.05 is considerably weaker than the standard threshold of 0.01, and no effective sample sizes or number of chains are reported. Given the wide, asymmetric posterior for c1 (best fit 9.85, mean 6.27 with 1σ lower bound 3.97) and the marginal detection of c3, the quoted 1σ intervals may not be converged. The claim of a preference for nonzero parameters should be verified with a stricter convergence criterion and with a report of effective sample sizes.","section":"Sec. V.A and Table I"},{"comment":"The prior ranges for c1, c2, and c3 are not specified. The text states that sampling starts around zero and that the parameters can take both positive and negative values 'without imposing strict bounds,' but the actual ranges used in the Monte Python runs are not given. This is essential for reproducibility and for interpreting the marginal posteriors, especially since wide or unbounded priors can slow convergence and influence the reported means and credible intervals.","section":"Sec. V.A"}],"minor_comments":[{"comment":"The statement that the equation-of-state singularity at ρϕ = 0 is 'physically acceptable' is asserted rather than demonstrated. Since the paper's central claim is the absence of instabilities, it would be useful to show explicitly that the perturbation variables, not just Qs and cs², remain finite through the crossing.","section":"Sec. IV.A and Fig. 2"},{"comment":"The '1σ preference' for c3 is marginal: Table I gives c3 = 0.00042 +0.00012/−0.00040, so the 68% interval excludes zero only at its lower edge. The wording in Sec. V.B ('preference for a positive, non-zero value of all the model parameters within 1σ') overstates the strength of the evidence; the authors should report the credible interval explicitly and temper the claim.","section":"Table I and Fig. 8"},{"comment":"The axis label '10+8c2' appears to be a typographical error for '10⁸ c2'; this should be fixed for clarity.","section":"Fig. 8"},{"comment":"The comparison with ΛCDM uses only Δχ² = −0.6. Since the model has three extra parameters, a model-selection criterion such as AIC or BIC would clarify whether the modest χ² improvement is penalized; this would also strengthen the statement that the model does not outperform ΛCDM.","section":"Sec. V.B"},{"comment":"The units of c2 are given as Mpc² and the paper states that c1 and c3 are dimensionless; this should be stated consistently in Table I, where the quoted quantity is 10⁻⁸c2, and in the prior description.","section":"Sec. III"}],"recommendation":"major_revision","confidential_remarks":"The manuscript addresses a timely topic and provides a useful existence proof for stable phantom crossing and negative dark energy density within Horndeski theory. The main concerns are technical: the sign-convention inconsistency for X is potentially load-bearing, and the data-driven preference for nonzero couplings is conditional on unvaried initial conditions and a loose convergence criterion. If the authors can resolve the sign convention issue, add robustness tests with varying φi and potential shapes, and tighten the MCMC diagnostics, the paper would be suitable for publication. I also recommend that the editors ask for the prior ranges to be reported explicitly."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does what it says on the tin: it gives a concrete Horndeski action whose background dynamics cross the phantom divide and yield negative dark energy density at high z, and it checks that scalar perturbations are free of ghosts and gradient instabilities. That part is clean. The new work relative to the authors' earlier paper is the full perturbation analysis and the combined MCMC constraints from Planck18, PantheonPlus, and BAO/f_sigma8. The fits are standard and honestly reported: Delta chi^2 = -0.6 against LCDM with three extra parameters, so the model does not resolve the H0 or S8 tensions in the joint analysis. The paper says so plainly. Credit is due for that honesty and for providing the stability conditions and the hi_class implementation details.\n\nWhere I part ways with the abstract's emphasis is the 'preference for non-zero values of the model parameters.' That preference is real only within the chosen convention phi_i = 10 at z~1000. The nonminimal coupling enters as c3*phi, so the physical strength of c3 is set relative to the field value. If you chose phi_i = 1, the best-fit c3 would shift by an order of magnitude, and the 1-sigma positive preference could easily disappear. The stress-test note is right about that. The paper does not sample or vary the initial conditions, nor does it test a different potential shape, so the claimed data-driven preference is conditional. It is not a fatal flaw—the existence proof is independent of this—but it does undercut the 'preference' language.\n\nTwo smaller issues. The Gelman-Rubin threshold R-1 < 0.05 is loose by modern standards; R-1 < 0.01 would be more convincing for a nine-parameter posteriors. And no chains or code are released, so the quantitative claims are not directly verifiable. These are fixable in revision, not reasons to reject.\n\nWho should read this: anyone working on modified-gravity dark energy or phantom crossing will find a useful worked example with explicit stability conditions. It is a legitimate data point in the 'can Horndeski do this without instabilities?' discussion. It deserves serious referee time; the authors should be asked to release chains and to show how the c3 posterior shifts under different phi_i choices. If that sensitivity is mild, the paper becomes considerably stronger. As it stands, accept for review with that request.","headline":"A solid existence proof for stable phantom crossing in a Horndeski model, but the data-driven preference for nonzero couplings is contingent on an unvaried field normalization and should be treated cautiously.","tokens_in":21617,"tokens_out":1860,"would_cite":true,"duration_ms":20500,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A Horndeski scalar-tensor dark energy model can cross the phantom divide and go negative at high redshift without instabilities, and combined cosmological data fit it as well as LambdaCDM.","keywords":["dynamical dark energy","Horndeski gravity","phantom crossing","negative energy density","cosmological tensions","scalar-tensor theory","growth of structure","Markov chain Monte Carlo"],"falsifier":"Recompute the background and perturbation evolution with the same $G_i$ functions but with initial conditions varied over, say, $\\phi_i\\in[1,20]$ and $\\phi'_i\\in[10^{-12},10^{-5}]$, retuning $V_0$ so that today's dark energy density is fixed: if the phantom crossing and negative high-redshift density disappear in any stable region of parameter space, the model's headline features are an artifact of the chosen initial data rather than a property of the Lagrangian.","tokens_in":20461,"feed_emoji":"🌌","tokens_out":11481,"duration_ms":94699,"temperature":0.7,"pith_summary":"The paper tries to establish that a specific Horndeski scalar-tensor model, in which dark energy is a scalar field nonminimally coupled to gravity and carrying derivative self-interactions, can reproduce two features that are impossible in general relativity without pathologies: the dark energy equation of state crossing the phantom divide $w=-1$, and a negative dark energy density at high redshifts. The authors show that both features arise from the dynamics of the field rather than from a phenomenological parameterization, while the kinetic self-interaction keeps the perturbations free of ghost and gradient instabilities. Confronting the model with Planck 2018 CMB data, BAO and $f\\sigma_8$ measurements, and PantheonPlus supernovae, they find a fit statistically comparable to $\\Lambda$CDM ($\\Delta\\chi^2 = -0.6$) and a $1\\sigma$ preference for positive values of all three coupling parameters. A sympathetic reader would care because stable phantom-like behaviour is exactly the kind of late-time modification often invoked to address the Hubble and $S_8$ tensions, and because recent large-scale-structure data have made dynamical dark energy a live question.","feed_headline":"Dark energy model crosses phantom divide without instabilities","feed_subtitle":"It fits CMB, BAO and supernova data as well as LambdaCDM while allowing negative dark energy at high redshift.","key_machinery":"The load-bearing object is the Horndeski Lagrangian with the three chosen functions $G_2 = X - V_0\\phi$, $G_3 = c_1\\phi + c_2X$, and $G_4 = \\frac{1}{2} + c_3\\phi$, with $G_5=0$ so that gravitational waves propagate at the speed of light. The nonminimal coupling $G_4 = \\frac{1}{2} + c_3\\phi$ produces the high-redshift negative energy density through the $-6c_3\\phi H^2$ term in $\\rho_\\phi$, while the derivative self-interaction $G_3$ controls late-time phantom behaviour; in particular $c_2$ maintains $Q_s>0$, so the phantom regime is reached without a ghost. The argument is carried by the Horndeski second-order action for perturbations, whose stability conditions $Q_s>0$ and $c_s^2>0$ the model is required to satisfy throughout the evolution.","core_discovery":"The central claim is that the Lagrangian $\\mathcal{L}_\\phi = \\frac{1}{2}\\partial_\\mu\\phi\\,\\partial^\\mu\\phi - V(\\phi) - (c_1\\phi + \\frac{1}{2}c_2\\,\\partial_\\mu\\phi\\,\\partial^\\mu\\phi)\\,\\Box\\phi + R(\\frac{1}{2}+c_3\\phi)$, with $V(\\phi)=V_0\\phi$, defines a Horndeski subclass (equivalently $G_2=X-V$, $G_3=c_1\\phi+c_2X$, $G_4=\\frac{1}{2}+c_3\\phi$, $G_5=0$) in which phantom crossing and negative dark energy at high redshift occur without instabilities. For positive $c_3$, the term $-6c_3\\phi H^2$ in the effective energy density dominates at early times, making $\\rho_\\phi$ negative while the total density stays positive; at low redshift the field's density turns positive and, for positive $c_1$, the equation of state enters the regime $w<-1$. The paper reports that the combined likelihood analysis of Planck 2018 CMB, BAO/$f\\sigma_8$, and PantheonPlus data gives $\\Delta\\chi^2 = -0.6$ relative to $\\Lambda$CDM, with posterior means $c_1 = 6.27^{+0.42}_{-2.3}$, $10^{-8}c_2 = 4.17^{+1.7}_{-3.0}$, and $c_3 = 0.00042^{+0.00012}_{-0.00040}$, all positive within $1\\sigma$. The model also predicts a suppressed growth rate at low redshift, but in the joint analysis it does not actually resolve the $H_0$ and $S_8$ tensions.","pith_inferences":["The paper's claim that 'this class of models' exhibits phantom crossing and negative densities is conditional on the fixed initial conditions $\\phi_i=10$, $\\phi'_i=10^{-10}$ at $z\\sim1000$ and the linear potential $V_0\\phi$; the paper does not test whether other initial data or potential shapes preserve these features, so the generality of the mechanism is unverified.","If negative high-redshift dark energy is real, it would suppress the early expansion rate and could feed structure formation; the model provides a concrete field-theoretic template for studying that effect, including its possible link to early massive galaxies.","A natural extension would be to rerun the same likelihood analysis with newer BAO data or with varied initial conditions; the $1\\sigma$ preference for nonzero couplings found here could sharpen or vanish, giving a sharp test of whether the negative-density feature is data-driven.","The contrast between the individual-dataset hints of tension relief and the joint-analysis null result suggests that other modified-gravity models claiming to resolve both tensions should be checked with full CMB perturbation likelihoods before conclusions are drawn."],"forward_implications":["If the model's stability and fit claims hold, a stable fundamental-field realization of phantom crossing exists, so late-time modifications of the expansion history need not be dismissed as ghost-ridden.","The statistically comparable fit to $\\Lambda$CDM means current data do not exclude this class of modified-gravity dark energy, and the weak preference for positive $c_3$ keeps the negative high-redshift density scenario observationally alive.","Because the joint analysis sharply limits the model's ability to raise $H_0$ or lower $S_8$, testing a modified-gravity dark energy model on perturbations is essential; background-only fits can overstate its tension-solving power.","Upcoming full large-scale-structure data releases and gravitational-wave speed measurements would discriminate the model's predictions for $w(z)$ and structure growth from $\\Lambda$CDM."],"supporting_citations":[{"why":"supplies the CMB temperature and polarization likelihood and the $\\Lambda$CDM baseline used for comparison.","marker":"[2]"},{"why":"provides the model-independent growth-rate measurement whose deviation from general relativity motivates the modified-gravity setup.","marker":"[4]"},{"why":"establishes that phantom crossing is a necessary condition for late-time models to simultaneously alleviate the $H_0$ and $S_8$ tensions, motivating the paper's construction.","marker":"[15]"},{"why":"shows that phantom crossing in general relativity generically leads to instabilities, defining the theoretical obstacle the Horndeski model must overcome.","marker":"[44]"},{"why":"supplies the perturbation-theory framework used to compute growth, stability conditions, and effective gravitational coupling in Horndeski gravity.","marker":"[49]"},{"why":"is the authors' earlier work introducing this model class and showing it can ease the $H_0$ tension, which the present paper extends to perturbations and joint data.","marker":"[52]"},{"why":"provides the numerical Boltzmann solver used to evolve background and perturbations and to enforce the stability conditions.","marker":"[55]"},{"why":"supplies the PantheonPlus supernova distance sample used in the likelihood.","marker":"[68]"}],"fun_headline_variants":["Horndeski model crosses phantom divide without instabilities","Negative dark energy at high z, stable in Horndeski gravity","Phantom crossing achieved stably in Horndeski dark energy","Horndeski DE fits data as well as LambdaCDM, no ghosts"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The phenomenology depends on fixed initial conditions for the scalar field at $z\\sim1000$, namely $\\phi_i=10$ and $\\phi'_i=10^{-10}$, and on the linear potential $V(\\phi)=V_0\\phi$ with $V_0$ tuned by a shooting method; these choices are not varied in the MCMC, so if different initial data or a different potential shape remove the phantom crossing or the negative-density epoch, the reported parameter preferences could change.","fun_headline_variants_meta":{"raw":{"variants":["Horndeski model crosses phantom divide without instabilities","Negative dark energy at high z, stable in Horndeski gravity","Phantom crossing achieved stably in Horndeski dark energy","Horndeski DE fits data as well as LambdaCDM, no ghosts"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000199,"raw_usage":{"total_tokens":1508,"prompt_tokens":1218,"completion_tokens":290,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":834,"completion_tokens_details":{"reasoning_tokens":213}},"tokens_in":834,"tokens_out":290,"duration_ms":3501,"temperature":1.0,"reasoning_tokens":213,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T04:51:37.745910+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the background and perturbation evolution with the same $G_i$ functions but with initial conditions varied over, say, $\\phi_i\\in[1,20]$ and $\\phi'_i\\in[10^{-12},10^{-5}]$, retuning $V_0$ so that today's dark energy density is fixed: if the phantom crossing and negative high-redshift density disappear in any stable region of parameter space, the model's headline features are an artifact of the chosen initial data rather than a property of the Lagrangian.","supporting_citations":[],"review_version":1}