{"id":"e5c516a5-635d-4b91-b144-694460ff8e62","arxiv_id":"2502.08670","paper_version":3,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"In a minimal turbulent channel, an injected wavelet-optimal resolvent forcing mode produces transient streamwise streaks that first match the linear optimal response, then break down via spanwise-advection-driven transfer into spanwise-doubled near-wall and streamwise-broken outer structures.","lead":"The authors inject an optimal time-localized forcing mode, computed from wavelet-based resolvent analysis, into a turbulent channel flow and track how the induced streaks grow and decay. The experiment shows that the linear optimal mode initially drives the streaks, but nonlinear effects, dominated by spanwise advection, clip the growth sooner in stronger forcing.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The frozen-mean body force F in Eq. (2.11) is load-bearing: the claimed early-time agreement between DNS and the linear response mode may be an artifact of artificially maintaining the base flow, and the authors' unfixed-mean check in §3.3 is asserted but not shown.","rationale":"I read the paper in good faith. The study is carefully executed: the DNS ensemble sizes are large (1000–4000), statistical convergence is checked by doubling the snapshot spacing, the wavelet resolvent formulation is built on prior published work, and the nonlinear energy-transfer analysis is detailed and physically consistent with known streak-breakdown mechanisms. The central claim is that the time-localised principal resolvent forcing mode transiently amplifies near-wall streaks in a fully nonlinear turbulent channel, that the DNS initially tracks the linear response mode, and that nonlinearity causes premature decay, with the principal mode outperforming random and suboptimal forcing. The weakest point in this argument is not the resolvent formalism itself but the experimental protocol: the mean profile is artificially frozen by the body force F, and the one control run that would test whether this freezing matters (an unfixed-mean simulation) is mentioned but not shown. The reader's verdict correctly identifies this as the load-bearing assumption. My read supports the same conclusion: the paper merits conditional acceptance, with the unfixed-mean check required before the quantitative agreement with the linear response can be taken as a robust property of the fully nonlinear system. The scaling-law fits to four points without error bars are a secondary concern and do not change the verdict. I therefore recommend no change to the reader's conditional verdict.","tokens_in":36468,"tokens_out":3061,"duration_ms":43985,"concrete_test":"Run the ε=5% forced DNS with the mean profile left free (i.e., without the F term in Eq. (2.11)), using the same ensemble of initial conditions and the same wavelet forcing mode. Compare the ensemble-averaged streak energy E^(0,1)(t) and the effective amplification σ_eff against the frozen-mean results in Fig. 3(a) and Fig. 4, and also report the time evolution of the mean-profile deviation from U1(y). If the unfixed-mean streak energy differs from the frozen-mean result by more than the reported statistical convergence bound (0.5–1.3%), or if σ_eff changes by more than a few percent, the early-time agreement with the linear response mode is not robust and the authors should provide the unfixed-mean data as a required supplement.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The resolvent modes are computed from the linearised Navier-Stokes equations about the unforced mean profile U1(y). In the forced DNS, Eq. (2.11) includes F=(F1,0,0), which removes the (0,0) contribution of the right-hand side at every time step, artificially pinning the mean to U1(y) throughout the simulation. Both the forced and the unforced baseline simulations use this same frozen-mean constraint, so the deviation operator Δ isolates the (0,±1) response while the base flow is prevented from reacting to the injected forcing. The linear response mode σ1ψ1 is the optimal response of the linearised operator about that same frozen U1, so the early-time agreement shown in Figs. 5 and 6 could be partly manufactured by the artificial body force: if the mean were free to evolve, the base flow would drift, the linear operator would no longer coincide with the resolvent operator, and the injected mode would no longer be optimally aligned with the instantaneous linear dynamics. The authors explicitly acknowledge this in §3.3, noting that 'allowing the mean profile to vary may reduce the effectiveness of the forcing mode', and they state that an unfixed-mean run for ε=5% gives a streak energy profile 'indeed very close' to Fig. 3(a). However, no figure, data, or quantitative comparison is provided for this check. Because the central claim—that the DNS response tracks the principal response mode before nonlinear breakdown—depends directly on the base flow remaining the one used to define optimality, this asserted but unshown control experiment is the most load-bearing unverified step in the paper.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper tests whether time-localised wavelet-based resolvent forcing modes, which are optimal for the linearised Navier-Stokes operator about a frozen mean profile, can transiently actuate near-wall streaks in a fully nonlinear minimal channel at Re_tau=186. Modes are computed by SVD of a windowed wavelet resolvent operator for the (k_x,k_z)=(0,1) Fourier mode, injected into an ensemble of forced DNS at amplitudes epsilon=1,2,5,10%, and compared with the optimal linear response mode. The authors report that the DNS tracks the optimal linear response at early times and near the wall, that nonlinearity causes premature decay with faster decay for stronger forcing, that the principal forcing mode outperforms random and second-suboptimal forcing, and that spanwise self-advection transfers energy to the (0,2) and (1,1) modes in the near-wall and outer regions. The manuscript also contains a quasi-linear model of the nonlinear energy transfer and resolvent analysis of the secondary modes.","tokens_in":36771,"tokens_out":7913,"duration_ms":67135,"significance":"If the central claims survive revision, the work provides a notable demonstration that transient resolvent modes can act as meaningful actuation structures in nonlinear turbulence and quantifies how nonlinearity clips linear optimal growth. The experimental design is careful in several respects: large ensembles (1000-4000), phase alignment via Eq. (2.10), modified wavenumbers matched to the DNS grid, and conservative comparisons against random and suboptimal forcing. The nonlinear energy transfer analysis and the quasi-linear model are valuable additions. The main caveat is that the frozen-mean constraint is load-bearing and the unfixed-mean check is asserted but not shown; this currently limits confidence in the claim that the DNS response tracks the linear response mode.","major_comments":[{"comment":"The artificial body force F that pins the mean streamwise profile to U1(y) is load-bearing for the central early-time agreement claim. The resolvent modes are optimal about this same frozen base flow, and both forced and unforced runs remove the (0,0) contribution of the right-hand side, so the comparison in Figs. 5-6 is between a constrained DNS and a linear operator defined about the same constrained base state. The authors acknowledge this and state in §3.3 that an unfixed-mean run for epsilon=5% gives a streak energy profile 'indeed very close' to Fig. 3(a), but no data, figure, or quantitative comparison is provided. Because allowing the mean to evolve would change the instantaneous linear operator and could reduce the alignment with the injected mode, the missing check directly affects the interpretation of the headline result. Please provide the unfixed-mean results (e.g., streak energy curve, peak and decay times, and the deviation from the linear response mode) or explicitly restrict the conclusions to the frozen-mean system.","section":"§2.4 (Eq. 2.11) and §3.3"},{"comment":"The scaling laws dE_hat_1 ~ |epsilon|^1.44 and dt_decay ~ |epsilon|^{-0.65} are presented as quantitative results but are fitted to only four forcing amplitudes with no uncertainty estimates, goodness-of-fit measures, or statement of the fit procedure. Since these exponents are used in the narrative that nonlinearities curtail linear growth and that stronger forcing accelerates decay, either provide confidence intervals and residual information or present these trends as qualitative descriptions of the four computed cases.","section":"§3.1 (Fig. 3b,d)"},{"comment":"The abstract's claim that the principal forcing mode is 'more effective' than the second suboptimal mode rests on the observation that sigma_eff for phi1 exceeds that for phi3 by only a factor of 1.03, while the corresponding linear ratio sigma1/sigma3 is 2.16. No confidence interval is given for this 3% difference. Given that the ensemble sizes vary between 1000 and 4000 and that the earlier convergence check (§2.4) was reported for streak energy rather than for sigma_eff, please provide a statistical uncertainty estimate (e.g., bootstrap across initial conditions) for the effective amplification ratio to support the ordering claim.","section":"§3.3 (Fig. 7)"}],"minor_comments":[{"comment":"The sentence 'such that the resolvent forcing mode is increasing the initial energy of the right-hand side by epsilon%' is grammatically unclear; 'increasing' should be 'increases' and the sentence should be rephrased.","section":"§2.4, text after Eq. (2.9)"},{"comment":"The sentence 'very close to the what is shown in figure 3(a)' contains a typo ('the what'); please correct it and, more importantly, reference a figure or table for the unfixed-mean result.","section":"§3.3, last paragraph"},{"comment":"The caption states 'epsilon≈5%' for the purple case, whereas all other cases use exact percentages; please state the exact value or explain the approximation.","section":"Figure 9 caption"},{"comment":"The sentence 'The energy content of the secondary modes (figures 8, 9) are the result of nonlinear interactions' has a subject-verb disagreement; 'are' should be 'is'.","section":"§4.1, first paragraph"},{"comment":"The random forcing mode phi_rand is not fully specified; please state how the random spatial field is sampled (e.g., Gaussian, wall-normal support) and how its normalization is performed, to enable reproducibility.","section":"§2.4, random forcing"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take: this is a well-executed study that tests whether a time-localised wavelet-based resolvent forcing mode actually actuates near-wall streaks in a fully nonlinear turbulent channel. The answer is yes for short times, but nonlinearities clip the growth. The new content is the systematic forced-DNS campaign—ensembles of 1000–4000, phase alignment, modified wavenumbers matched to the DNS grid—and the breakdown analysis identifying (0,2) and (1,1) secondary modes with spanwise self-advection dominating. That part is solid. I agree with the reader that the central claims hold up.\n\nSoft spots: the frozen-mean body force F in Eq. (2.11) is indeed load-bearing. The resolvent modes are optimal for that frozen profile, and the early-time comparison to the linear response is only meaningful if the base flow stays fixed. The authors state in §3.3 that an unfixed-mean run gives nearly identical streak energy, but they don't show it. Given that this is the one control that would validate the comparison, not showing it is a genuine gap. That said, the forcing is short-lived (about 1.5 h/u_tau) and the linear growth happens before the mean can respond much, so I don't think it's fatal. It's a caveat that should be addressed in revision, not a reason to reject. The scaling laws (peak energy ~ ε^1.44, decay time ~ ε^-0.65) are fitted to four points with no error bars. That's worth flagging, but it's a minor issue—the qualitative trend is clear.\n\nI also want to disagree slightly with any claim that the early-time agreement is manufactured. The phase alignment and the use of unforced baseline simulations are careful. The fact that the response diverges from linear at different times for different amplitudes is consistent with genuine nonlinear breakdown, not an artifact.\n\nBottom line: this is a solid contribution for people working on resolvent analysis, flow control, or streak dynamics. It deserves a serious referee and likely conditional acceptance after the unfixed-mean check is shown and the scaling fits are quantified. I'd cite it if I were working on transient growth of resolvent modes. Bring it to reading group.","headline":"Careful forced-DNS test of wavelet resolvent modes shows early linear agreement then nonlinear decay; frozen-mean caveat is real but not fatal.","tokens_in":37341,"tokens_out":1902,"would_cite":true,"duration_ms":17707,"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":"Injected linear mode briefly drives near-wall streaks in turbulent channel flow.","keywords":["resolvent analysis","wavelet transform","turbulent channel flow","near-wall streaks","transient growth","nonlinear energy transfer","minimal flow unit","flow control"],"falsifier":"Run the same forced DNS with the mean-fixing body force removed at $\\varepsilon = 10\\%$; if the DNS still tracks the linear response mode up to $t\\approx 0.7\\,h/u_\\tau$ and the peak streak energy still scales sub-quadratically, the frozen-mean assumption is not biasing the comparison, but if the deviation begins earlier or the peak shifts significantly, the paper's attribution of the premature decay to nonlinearity is called into question.","tokens_in":36234,"feed_emoji":"🌊","tokens_out":6703,"duration_ms":56926,"temperature":0.7,"pith_summary":"This paper asks whether a resolvent forcing mode computed from the linearized Navier-Stokes equations, localized in time as a wavelet pulse, can actually actuate the near-wall cycle of a fully turbulent channel. The answer is a qualified yes: when injected into a minimal flow unit at $Re_\\tau\\approx 186$, the principal mode (streamwise rolls) drives streamwise streaks that grow and then decay, and the DNS tracks the optimal linear response mode at early times and near the wall. Nonlinearity cuts that growth short, stronger forcing causes faster decay, and the linear mode overpredicts the achievable amplification. The paper then identifies the breakdown mechanism: energy is transferred from the forced streak to a spanwise-doubled mode near the wall and a streamwise-modulated mode in the outer region, dominated by spanwise self-advection. If this is right, linear resolvent modes remain cheap and useful as actuators for control-oriented prediction, but only with a built-in nonlinearity penalty.","feed_headline":"Injected linear mode briefly drives near-wall streaks","feed_subtitle":"The optimal wavelet pulse grows transient streaks in turbulent flow, but nonlinear decay starts sooner with stronger forcing.","key_machinery":"The central object is the windowed wavelet-based resolvent operator $\\tilde{H}^{(0,1)} B$, where $\\tilde{H}$ is the discrete linearized Navier-Stokes operator in a wavelet-in-time basis and $B$ is a windowing matrix that restricts the forcing to a compact Daubechies-8 scaling-function pulse at a chosen scale and shift. An SVD of this combined operator yields the optimal forcing mode (streamwise rolls) and the corresponding transient response mode (streamwise streaks), ordered by time-integrated kinetic-energy amplification. The second half of the machinery is the scale-to-scale energy-transfer diagnostic $\\hat{T}^{(0,1)}_{(p_1,p_3)}$, which decomposes nonlinear transfer from the actuated mode by interacting wavenumbers and by advection direction; this diagnostic localizes the $(0,2)$ and $(1,1)$ sinks and identifies spanwise self-advection as the dominant pathway.","core_discovery":"The central claim is that the time-localized principal resolvent forcing mode, obtained from an SVD of the windowed wavelet-based resolvent operator, is an effective but imperfect actuator for the buffer-layer streak cycle. In the minimal flow unit at $Re_\\tau\\approx 186$, forcing that mode at intensities $\\varepsilon = 1\\%$ to $10\\%$ of the unforced nonlinearity produces the expected rolls-to-streaks lift-up growth, and the instantaneous streamwise velocity deviation collapses onto the linear response mode for $t \\lesssim 0.7\\,h/u_\\tau$ and $y^+ \\lesssim 15$. Beyond that, the response decays prematurely in all cases, with peak streak energy scaling sub-quadratically and decay time scaling as $\\varepsilon^{-0.65}$. The principal mode still outperforms the first suboptimal mode and a random forcing structure at amplifying near-wall streaks, though the effective amplification gap is small. The paper also claims that the nonlinear breakdown follows a fixed spatial template: spanwise self-advection feeds a $(0,2)$ mode in the near-wall region and a $(1,1)$ mode in the outer region, and resolvent modes of those secondary scales sit exactly at the foci of energy transfer.","pith_inferences":["One editorial extension: the same wavelet-resolvent construction could be applied to streak modes of different spanwise wavelengths to test whether linear amplification or nonlinear transfer controls streak spacing in larger channels.","The paper's frozen-mean setting is its most delicate assumption; an unfixed-mean experiment at $\\varepsilon = 10\\%$ would cleanly check whether the early-time agreement is an artifact of holding the base flow constant.","A practical control corollary left implicit is that there is a sweet spot in forcing intensity: too weak actuation wastes the mode's efficiency, while too strong trips the fast spanwise-advection decay, so an intermediate $\\varepsilon$ should maximize time-integrated streak energy."],"forward_implications":["Resolvent-based control designs must discount linear amplification by an intensity-dependent factor, because the DNS amplification coefficient $\\sigma_{eff}$ is always below $\\sigma_1 = 11.54$ and decreases with forcing strength.","Streak breakdown is wavenumber-selective: actuating the $(0,1)$ mode feeds the $(0,2)$ mode near the wall and the $(1,1)$ mode in the outer layer, so control aimed at sustaining streaks should target those two secondary modes.","The dominance of spanwise self-advection in the nonlinear energy transfer implies that reducing spanwise gradients of the actuated streak, rather than streamwise or wall-normal coupling, is the likeliest way to prolong transient growth.","The early collapse of the DNS onto the linear response mode, lasting roughly an eddy turnover time near the wall, supports using the wavelet-resolvent mode as a short-horizon predictor of actuation effects in wall-bounded turbulence."],"supporting_citations":[{"why":"Supplies the wavelet-based resolvent formulation and the SVD from which the modes are computed.","marker":"Ballouz et al. 2024b"},{"why":"Established that principal resolvent forcing modes contribute to the self-sustaining buffer-layer process and provides the baseline mode shapes and wavenumbers.","marker":"Bae et al. 2021"},{"why":"Defines the minimal flow unit used for the DNS experiments.","marker":"Jiménez & Moin 1991"},{"why":"Provides the nonlinear energy-transfer decomposition used to quantify scale-to-scale transfer.","marker":"Symon et al. 2021"},{"why":"Extends the mode-to-mode nonlinear energy-transfer diagnostic applied in the breakdown analysis.","marker":"Ding et al. 2025"},{"why":"Shows spanwise-gradient-driven transient growth of secondary modes, used to support the spanwise-advection dominance finding.","marker":"Markeviciute & Kerswell 2024"},{"why":"Supports the argument that linear transient growth sustains minimal-channel turbulence, the premise of the actuation test.","marker":"Lozano-Durán et al. 2021"}],"fun_headline_variants":["Wavelet pulse spurs fleeting streaks in turbulence","Optimal linear mode excites streaks, then fades","Transient streak growth from a single resolvent pulse","Nonlinearity curbs linear mode's streak boost","Wavelet-based mode yields short-lived near-wall streaks"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole comparison assumes the mean streamwise profile stays exactly as in the unforced flow, enforced by an artificial body force, so the resolvent mode remains optimal for the base flow the DNS actually sees.","fun_headline_variants_meta":{"raw":{"variants":["Wavelet pulse spurs fleeting streaks in turbulence","Optimal linear mode excites streaks, then fades","Transient streak growth from a single resolvent pulse","Nonlinearity curbs linear mode's streak boost","Wavelet-based mode yields short-lived near-wall streaks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000573,"raw_usage":{"total_tokens":2787,"prompt_tokens":1107,"completion_tokens":1680,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":723,"completion_tokens_details":{"reasoning_tokens":1614}},"tokens_in":723,"tokens_out":1680,"duration_ms":10290,"temperature":1.0,"reasoning_tokens":1614,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T11:02:21.448379+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same forced DNS with the mean-fixing body force removed at $\\varepsilon = 10\\%$; if the DNS still tracks the linear response mode up to $t\\approx 0.7\\,h/u_\\tau$ and the peak streak energy still scales sub-quadratically, the frozen-mean assumption is not biasing the comparison, but if the deviation begins earlier or the peak shifts significantly, the paper's attribution of the premature decay to nonlinearity is called into question.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Established that principal resolvent forcing modes contribute to the self-sustaining buffer-layer process and provides the baseline mode shapes and wavenumbers."},{"cited_title":", Illingworth, S","cited_arxiv_id":null,"evidence_quote":"Provides the nonlinear energy-transfer decomposition used to quantify scale-to-scale transfer."},{"cited_title":"Journal of Fluid Mechanics 1002 , A42","cited_arxiv_id":null,"evidence_quote":"Extends the mode-to-mode nonlinear energy-transfer diagnostic applied in the breakdown analysis."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows spanwise-gradient-driven transient growth of secondary modes, used to support the spanwise-advection dominance finding."}],"review_version":1}