{"id":"130dcfd5-ce80-4284-b210-c11500279df9","arxiv_id":"2411.16449","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The alpha-to-delta transition in the Jansen-Rit neural mass model is identified as a grazing bifurcation at which the pyramidal-cell potential touches the excitatory-interneuron switch-off threshold.","lead":"This paper shows that a sharp switch in brain-wave rhythms in a standard model of a cortical column is a grazing event, where one signal just barely touches a threshold and excitatory feedback abruptly switches off. The result gives a mathematical way to predict where alpha waves give way to delta waves in the model, which is relevant to sleep and brain-state research.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Grazing curve is derived in a limit where delta oscillations do not exist, so the alpha-delta identification rests on numerical overlap; the paper's own Section 4 admission makes this a genuine gap.","rationale":"The paper does several things well: it gives an explicit algebraic construction of alpha-type periodic orbits and their grazing condition, it provides reproducible scripts, and Figure 6 shows that the grazing curve tracks the finite-epsilon alpha-delta boundary closely, including the SNP folds and the frequency drop. These are genuine, independent supports for a weaker claim: grazing is the singular-limit mechanism that ends alpha-type oscillations and initiates the near-zero phase that characterizes delta at physiological epsilon. However, the abstract's stronger formulation, 'we identify the transition between alpha and delta oscillations as a discontinuity-induced grazing bifurcation,' requires the chosen singular limit to contain both sides of the transition. It does not. The reader's weakest assumption identifies exactly this: strategy (a) is the limit in which delta oscillations do not exist, so the accuracy of the grazing curve as the transition boundary is only numerically demonstrated, not derived. The paper's own Section 4 confirms the limitation, and the discussion of strategy (b) makes clear that a complete singular perturbation analysis of the delta side is absent. This is a serious but addressable caveat: a convergence study of the SNP/frequency-drop boundary as epsilon decreases, together with a matched analysis for thresholds proportional to epsilon, would settle whether the grazing curve is the true asymptotic boundary or merely a good numerical approximation at epsilon = 0.024. The verdict CONDITIONAL is therefore appropriate; the concern does not invalidate the paper but does mean the central claim currently rests on an unproved identification of limits. My analysis does not change the reader's verdict, so I mark the verdict UNCHANGED.","tokens_in":20849,"tokens_out":7147,"duration_ms":68368,"concrete_test":"Perform a convergence study for the full system (5) at b* = 0.4 and b* = 0.3: for epsilon = 0.024, 0.012, 0.006, and 0.003, continue the saddle-node-of-limit-cycle fold (or, where absent, the sharp frequency-drop contour) that marks the alpha-delta boundary, and track its location in G. Then compare the extrapolated epsilon -> 0 limit of this curve with the strategy-(a) grazing curve from (20)-(21), and also with a grazing curve recomputed under strategy (b), where y0,1 and y0,3 are taken proportional to epsilon. In parallel, at a point below the grazing curve (for example G = 1.6, b* = 0.4), integrate (5) for decreasing epsilon and measure the delta orbit's minimum y1 and period; if the minimum tends to zero and the period diverges, delta orbits vanish in limit (a), confirming the mismatch.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires the grazing curve from equations (20)-(21) to be the epsilon-to-zero limit of the alpha-delta transition. But the analyzed limit, strategy (a) of Section 3.2, keeps the excitatory thresholds y0,1 = y0,3 = 0.08 fixed as epsilon tends to zero. In this limit the off equilibrium yeq1 = (0,0,0) is stable for all G (Table 3), so after grazing the trajectory collapses to rest; no delta-type periodic orbit exists in the limit. The paper states this explicitly in Section 4: 'the limit epsilon -> 0 with fixed thresholds y0,1, y0,3 does not support delta-type oscillations.' The grazing curve is therefore rigorously a boundary of the alpha branch only, not a bifurcation between two coexisting oscillation types in the same singular limit. The finite-epsilon transition is in fact bounded by a pair of saddle-node-of-limit-cycle folds (SNP, Figure 6a), and the claim that these folds 'overlap with the grazing bifurcation' is an observation at epsilon = 0.024, not a derived asymptotic statement. Strategy (b), where the excitatory thresholds scale with epsilon and delta activity does exist, is discussed but not used to derive a grazing curve. Thus the identification of the alpha-delta transition as a grazing bifurcation is weaker than the abstract states; what is firmly established is that grazing marks the collapse of alpha-type support, which in the finite-epsilon system is followed by the slow near-zero phase of delta.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the Jansen-Rit neural mass model and the sharp transition between alpha- and delta-type oscillations. After non-dimensionalisation, a small parameter epsilon is identified, and in the singular limit epsilon -> 0 the sigmoidal activation functions become Heaviside switches. Using strategy (a) (holding the excitatory thresholds y0,1 and y0,3 fixed at 0.08), the authors derive explicit formulas for equilibria (Section 3.3) and an algebraic system, equations (20)-(21), for alpha-type periodic orbits and their grazing bifurcation, where the minimum of the pyramidal-cell output y1 equals the threshold y0,3. They track the grazing curve in the (b*, G)-plane and compare it with numerical bifurcation data at the physiological value epsilon = 0.024 (Figure 6), concluding that the alpha-delta transition is a grazing bifurcation.","tokens_in":21207,"tokens_out":8831,"duration_ms":77101,"significance":"If the central claim is appropriately qualified, the paper makes a valuable contribution. The algebraic derivation of the grazing curve in the singular limit is a substantial piece of analysis, and the explicit formulas for equilibria in Section 3.3 are useful. The paper is commendably transparent: it ships reproducible computational scripts, and the derivation involves no free parameters fit to the target transition. The identification of the alpha-delta transition as a grazing phenomenon is novel and could stimulate further work on discontinuity-induced bifurcations in neural mass models. However, the central claim is only partially supported by the analysis: the singular limit used for the grazing curve does not contain delta-type oscillations, and the numerical comparison with the finite-epsilon system is made at a single value of epsilon. These gaps require revision before the paper can be accepted.","major_comments":[{"comment":"The abstract states that the paper identifies the transition between alpha and delta oscillations as a grazing bifurcation, and the Figure 6 caption calls the grazing curve 'Asymptotically for epsilon -> 0 the transition boundary between delta and alpha activity'. This overstates what is established. The singular limit used for the derivation is strategy (a), with y0,1 = y0,3 = 0.08 fixed. In this limit delta-type oscillations do not exist: Section 4 explicitly states that 'the limit epsilon -> 0 with fixed thresholds y0,1, y0,3 does not support delta-type oscillations', and Table 3 shows that the stable 'off' equilibrium yeq1 = (0,0,0) is present for all G. Therefore the grazing curve from equations (20)-(21) is rigorously a boundary of the alpha-type periodic orbit branch only; it is not a bifurcation between two coexisting oscillation types in that limit. The alpha-delta identification at epsilon = 0.024 rests on the numerical overlap of the grazing curve with the SNP folds in Figure 6a, which is an observation, not a derived asymptotic statement. The abstract and introduction should be revised to state that grazing marks the collapse of alpha-type support, and that for the smooth system at physiological epsilon this collapse coincides with the numerically observed alpha-delta transition.","section":"Section 4 (Small-threshold analysis) and abstract"},{"comment":"The quantitative claim of accuracy is not backed by a convergence study or an error bound. The SNP folds are computed only at epsilon = 0.024, and the text itself acknowledges that 'for larger b* a small deviation is noticeable'. To support the statement in the introduction that the grazing bifurcation is an 'accurate approximation' of the boundary between alpha- and delta-type oscillations, the paper should report the distance between the grazing curve and the SNP folds (for example, the maximal error in G or b* over the plotted parameter range), and ideally show the SNP curves for a sequence of smaller epsilon values approaching the epsilon -> 0 grazing curve.","section":"Section 3.4, Figure 6a"},{"comment":"The statement in Section 3.2 that 'In both limits the collapse of the alpha-frequency oscillations is a grazing bifurcation' is not derived for strategy (b), in which y0,1 and y0,3 scale with epsilon and in which delta-type oscillations exist. The analysis of strategy (b) is limited to showing that the small-activity equilibria disappear (Section 3.3, 'Brief comment on the excitatory activation thresholds'), and Section 4 defers a detailed analysis to future work. Since strategy (b) is the limit that actually supports delta oscillations, the paper should either derive or numerically demonstrate the grazing condition in that threshold-scaling limit, or explicitly mark the assertion as a conjecture.","section":"Section 3.2 and Section 4"}],"minor_comments":[{"comment":"The units of y0 are given as '[mV]−1'; since y0 is an activation threshold, the units should be 'mV', not 'mV^-1'.","section":"Section 2, Table 1"},{"comment":"The representative points in the alpha and delta regions are given as '(0.4, 0.14)' and '(0.4, 0.16)' in the body text, but should be '(0.4, 1.4)' and '(0.4, 1.6)' as in the figure caption.","section":"Section 3.4, text near Figure 6"},{"comment":"The abbreviation 'eg.' should be 'e.g.'.","section":"Abstract"},{"comment":"The phrase 'exictatory interneurons' is a typo for 'excitatory interneurons'.","section":"Section 2"},{"comment":"The definition of Rhs0(y1) would be clearer if the three cases for the different G-intervals (as given in the table below) were included explicitly, since the single conditional expression does not display the G-dependence of the right-hand side.","section":"Equation (12)"},{"comment":"The frequency shading is described in the text but the figure does not contain a color bar or legend explaining the frequency scale; adding one would aid interpretation.","section":"Figure 6a"}],"recommendation":"major_revision","confidential_remarks":"The paper contains a genuinely interesting and mostly sound mathematical analysis, and the authors are transparent about the limitation of the singular limit in Section 4. My main concern is that the abstract and introduction do not reflect this limitation, and the numerical evidence for the alpha-delta identification is limited to a single value of epsilon. With careful reframing of the central claim and an explicit error or convergence statement, the paper would be a solid contribution to the dynamical-systems analysis of neural mass models."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Fairly good paper. The main new thing: they show that the sharp alpha-delta transition in the Jansen-Rit model can be located as a grazing bifurcation in a singular limit, and they give explicit piecewise-exponential formulas for alpha-type periodic orbits (system 20-21). Prior work by Forrester et al. called it a 'false bifurcation' and tracked an inflection point; this paper gives a mechanistic interpretation and an algebraic detection method. The numerical comparison at the physiological epsilon = 0.024 is honest: the grazing curve overlaps the frequency drop and the fold curves over a large parameter range. Code and data are available, and there is no parameter fitting to the target transition. That is real evidence.\n\nThe soft spot is real but the authors see it. The singular limit they analyze (strategy (a)) keeps the excitatory thresholds y0,1,y0,3 fixed O(1) as epsilon -> 0. In that limit the 'off' equilibrium is stable and delta-type oscillations do not exist. So the grazing curve is rigorously a boundary of the alpha branch, not a bifurcation between two oscillation types in the same limit. The claim that grazing is the alpha-delta transition rests on numerical overlap at finite epsilon plus the sensible mechanism that collapse of the excitatory feedback initiates the long near-zero phase of delta. The paper's own Section 4 says the limit does not support delta oscillations and discusses strategy (b) (thresholds scaling with epsilon) where delta exists, but they don't derive a grazing curve there. So the abstract overstates the result a bit. This is a serious-but-addressable gap, not a fatal flaw.\n\nMinor issues: the writing has a few typos and awkward phrases, but nothing that obscures the math. The canard explosion near the Hopf point is intriguing but not developed; it's a side comment, not load-bearing.\n\nWho should read this: people using neural mass models for EEG rhythms, and anyone working on grazing bifurcations in piecewise-smooth systems. It's a good candidate for peer review. As a referee I'd push for either a more careful statement of what is derived in the limit versus observed at finite epsilon, or an extension to the small-threshold limit. The paper's own Section 4 gives the roadmap.","headline":"Solid, useful paper: identifies the alpha-delta transition in the Jansen-Rit model as a grazing bifurcation with explicit formulas, but the singular limit used doesn't support delta oscillations, so the central claim rests partly on numerical overlap; still deserves peer review.","tokens_in":21650,"tokens_out":2979,"would_cite":true,"duration_ms":26706,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["34A36","34C23","37G15","37N25"],"pacs":[],"model":"deepseek-v4-flash","headline":"The sharp switch from alpha to delta rhythm in the Jansen-Rit cortical-column model is a grazing bifurcation: the orbit's minimum touches the threshold that cuts excitatory feedback, collapsing activity until slow inhibition recovers.","keywords":["Brain activity","Neural mass model","Alpha and Delta rhythms","Bifurcation analysis","Heaviside function","Piecewise smooth dynamical systems","grazing bifurcation","Jansen-Rit model"],"falsifier":"For a sequence of decreasing $\\epsilon$ values with the excitatory thresholds held fixed, continue the $\\alpha$-type periodic orbits of the full smooth system (5) and record the parameter values where the minimum of $y_1$ first equals $y_{0,3}$; the grazing mechanism is confirmed if this locus converges to the curve given by (20)-(21) as $\\epsilon \\to 0$. If the locus diverges, or if the sharp frequency drop disappears from the neighbourhood of the curve for any $\\epsilon$ smaller than 0.024, the claim would be refuted.","tokens_in":20657,"feed_emoji":"🧠","tokens_out":13063,"duration_ms":103620,"temperature":0.7,"pith_summary":"This paper aims to explain a sharp transition observed in the Jansen-Rit model, a widely used neural-mass model of a single cortical column, between fast $\\alpha$-type oscillations near 10 Hz and slow delta-type oscillations near 2-4 Hz. The transition occurs over a tiny parameter range with little or no hysteresis, which rules out the standard bifurcations of smooth dynamical systems as the underlying mechanism. The authors non-dimensionalise the model and observe that the excitatory activation functions are steep with small thresholds, so a singular limit in which they become all-or-nothing Heaviside switches is appropriate. In that limit they derive an algebraic system for the $\\alpha$-type periodic orbits, locate the grazing bifurcation where the minimum of the pyramidal-cell output equals the threshold for switching off the excitatory interneurons, and show numerically that this grazing curve tracks the $\\alpha$-delta boundary at the physiological value $\\epsilon = 0.024$. A reader should care because the paper gives a mechanistic account of a brain-state switch relevant to sleep and emotional regulation, replacing an earlier purely descriptive identification of the transition.","feed_headline":"Cortex's alpha-delta switch is a threshold-touch event","feed_subtitle":"A cortical-column model's fast alpha rhythm collapses the moment a feedback threshold is grazed.","key_machinery":"The argument rests on a piecewise-linear reduction of the non-dimensionalised Jansen-Rit model, obtained by letting the small parameter $\\epsilon$ (a quarter of the inverse activation slope) tend to zero so that each sigmoid becomes a Heaviside switch. In this limit the $\\alpha$-type orbit is a piecewise-exponential solution of two coupled affine second-order oscillators, (16)-(17), for the pyramidal-cell activity $y_1$ and the inhibitory activity $y_2$, with the excitatory interneuron activity $y_3$ pinned at its equilibrium value $2\\alpha_2/G$. The paper parametrises such orbits by their four switching times using the explicit solution formula for the affine oscillator $\\ddot{y} = b^2 c - 2b\\dot{y} - b^2 y$, which yields the four algebraic equations (20); adjoining the grazing condition (21), which sets the minimum of $y_1$ equal to $y_{0,3}$ with vanishing time derivative, turns the grazing bifurcation into a root-finding problem in two parameters. This finite-dimensional encoding is what makes the transition boundary computable and trackable.","core_discovery":"The central discovery is that, in the singular limit where the excitatory activation functions of the model become Heaviside switches, the $\\alpha$-delta transition is a discontinuity-induced grazing bifurcation of the $\\alpha$-type periodic orbit. During $\\alpha$ oscillations the pyramidal-cell potential $y_1$ stays above the threshold $y_{0,3}$ at which the excitatory interneurons switch off; at the transition, the orbit's minimum $y_{1,\\min}$ grazes this threshold in a quadratic tangency, the excitatory interneuron activity collapses, and both excitatory populations drop to near-zero values, producing the relaxation-type profile characteristic of delta activity. The paper encodes this condition as an extra algebraic equation, $y_{1,\\min} = y_{0,3}$, appended to the switching-time system (20), and tracks the resulting curve in the $(b^*, G)$ parameter plane, where $b^*$ is the ratio of inhibitory to excitatory decay rates and $G$ is the ratio of inhibitory to excitatory feedback strengths. This grazing curve separates $\\alpha$ from delta activity and coincides, over large parts of the plane, with the pair of saddle-node-of-limit-cycle folds and the sharp frequency drop computed for the full smooth system at $\\epsilon = 0.024$. The paper notes that this fixed-threshold limit does not itself support delta oscillations, which require the alternative limit in which the excitatory thresholds scale to zero with $\\epsilon$.","pith_inferences":["If the grazing mechanism is correct, the alpha-delta boundary should be visible in any smooth model with the same small thresholds: as the activation slope increases, the numerically detected tangency of the orbit to the threshold should approach the curve (20)-(21), a testable prediction that the paper checks at only one slope value.","The two singular limits the authors contrast, fixed thresholds versus thresholds of order $\\epsilon$, correspond to two different physiological regimes: a collapse into an off state versus genuine slow oscillations. An implicit consequence is that real delta activity requires excitatory thresholds comparable to the noise floor, so a cortical measurement of effective activation thresholds could dec","The grazing condition is purely geometric, so the same mechanism should persist under moderate changes in connectivity parameters as long as the excitatory thresholds stay small; this suggests the alpha-delta switch is robust across parameter variations, a point the paper does not explore."],"forward_implications":["The grazing curve computed from (20)-(21) is the singular-limit boundary between alpha and delta activity, and at the physiological value $\\epsilon = 0.024$ it coincides with the numerically observed sharp frequency drop in the $(b^*, G)$ plane.","Because the transition is a tangency rather than a stability change, the mechanism explains why the alpha-delta switch appears over a small parameter range with almost no hysteresis.","In the fixed-threshold singular limit, a trajectory that grazes the threshold collapses to the stable 'off' equilibrium; genuine relaxation-type delta oscillations require the alternative limit in which the excitatory thresholds shrink to zero together with $\\epsilon$.","The formula for the Hopf bifurcation at $G_1$ in the singular limit gives an explicit prediction for the onset of alpha activity, and the alpha orbits near it display canard-like explosive growth whose parameter-amplitude dependence is in principle computable from the piecewise-exponential formulas."],"supporting_citations":[{"why":"supplies the original cortical-column model whose alpha-delta transition is the object of study.","marker":"[18]"},{"why":"gives the baseline bifurcation analysis of the model that identifies the oscillatory regimes this paper re-examines.","marker":"[15]"},{"why":"provides the non-dimensionalisation of the model that reveals the small parameter $\\epsilon$.","marker":"[27]"},{"why":"previously identified the transition as a 'false bifurcation' that this paper replaces with a grazing mechanism.","marker":"[13]"},{"why":"provides the theory of grazing bifurcations and piecewise-smooth dynamics that names the mechanism.","marker":"[11]"},{"why":"defines the pseudo-equilibria framework used to classify equilibria of the piecewise-linear limit.","marker":"[22]"},{"why":"grounds the assumption that excitatory populations have steeper activation slopes and smaller thresholds than the inhibitory population.","marker":"[31]"},{"why":"supplies the continuation toolbox used to compute the bifurcation diagrams and to verify the grazing curve.","marker":"[8]"}],"fun_headline_variants":["Alpha-delta transition is a grazing bifurcation","Grazing bifurcation triggers alpha-delta rhythm shift","Cortical alpha-delta switch: a threshold-grazing event","Threshold grazing flips alpha rhythm to delta","Feedback collapse at threshold marks alpha-delta switch"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the real model's sharp transition, where activation thresholds are small but not zero, is well approximated by the idealized limit of infinitely steep on-off switches, a limit in which genuine slow delta oscillations do not actually exist, so the match is demonstrated only numerically at one parameter value.","fun_headline_variants_meta":{"raw":{"variants":["Alpha-delta transition is a grazing bifurcation","Grazing bifurcation triggers alpha-delta rhythm shift","Cortical alpha-delta switch: a threshold-grazing event","Threshold grazing flips alpha rhythm to delta","Feedback collapse at threshold marks alpha-delta switch"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000664,"raw_usage":{"total_tokens":3079,"prompt_tokens":1042,"completion_tokens":2037,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":658,"completion_tokens_details":{"reasoning_tokens":1961}},"tokens_in":658,"tokens_out":2037,"duration_ms":16082,"temperature":1.0,"reasoning_tokens":1961,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:06:15.150437+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a sequence of decreasing $\\epsilon$ values with the excitatory thresholds held fixed, continue the $\\alpha$-type periodic orbits of the full smooth system (5) and record the parameter values where the minimum of $y_1$ first equals $y_{0,3}$; the grazing mechanism is confirmed if this locus converges to the curve given by (20)-(21) as $\\epsilon \\to 0$. If the locus diverges, or if the sharp frequency drop disappears from the neighbourhood of the curve for any $\\epsilon$ smaller than 0.024, the claim would be refuted.","supporting_citations":[{"cited_title":"Grimbert and O","cited_arxiv_id":null,"evidence_quote":"gives the baseline bifurcation analysis of the model that identifies the oscillatory regimes this paper re-examines."},{"cited_title":"Codimension Two Bifurcations and Rythms in Neural Mass Models","cited_arxiv_id":"0907.2718","evidence_quote":"provides the non-dimensionalisation of the model that reveals the small parameter $\\epsilon$."},{"cited_title":"Forrester, J","cited_arxiv_id":null,"evidence_quote":"previously identified the transition as a 'false bifurcation' that this paper replaces with a grazing mechanism."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"defines the pseudo-equilibria framework used to classify equilibria of the piecewise-linear limit."}],"review_version":1}