REVIEW 3 major objections 5 minor 55 references
New islands of stability with double-trace deformations
T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read Fully resonant Neumann boundary conditions still admit stable, strongly multi-mode oscillators arbitrarily close to AdS, contradicting the single-mode-dominated picture of the islands of stability.
desk verdict A credible numerical discovery—Neumann AdS admits strongly multi-mode non-collapsing solutions near AdS—but the 'arbitrarily close' claim rests on extrapolation from epsilon=0.001, so treat that part as suggestive. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central object is the double-oscillator, a quasi-periodic solution oscillating on two frequencies, constructed by imposing a time-periodic Fourier ansatz and solving the Einstein–Klein–Gordon equations as a boundary value problem with a Newton–Raphson method and pseudo-spectral discretization. The boundary-condition family is the Robin condition $\sin(\pi\kappa/2)\varphi_1-\cos(\pi\kappa/2)\varphi_2=0$, with $\kappa=1$ giving the fully resonant Dirichlet spectrum (frequencies $\omega_n=2n$) and $\kappa=0$ giving the fully resonant Neumann spectrum (frequencies $\omega_n=2n+1$). Resonance is diagnosed through frequency quadruples satisfying $\Delta\omega_J=\omega_{j_2}+\omega_{j_3}-\omega_{j_4}-\omega_{j_1}=0$. The key diagnostic is the projection of a solution onto the normal modes; the exponential decay rate of those mode amplitudes distinguishes two-mode-dominated from strongly multi-mode data.
What would settle it
Repeat the construction at $\varepsilon=10^{-4}$ or smaller and check whether the branch persists with a nonzero exponential mode-decay rate; or evolve the $\kappa=0$ double oscillator with $\varepsilon=0.001$ for times much longer than $\varepsilon^{-3}$ and watch whether $\varphi\varphi^*$ at the boundary remains periodic or develops drift toward horizon formation. Failure of either test would disqualify the 'arbitrarily close to AdS' claim.
Extended reading notes
Core claim
The paper claims that the fully resonant Neumann spectrum admits islands of stability containing strongly multi-mode data arbitrarily close to AdS, and that these solutions are not single-mode dominated. Concretely, for a complex scalar field with $m^2=-2$, the authors construct double-oscillators, quasi-periodic solutions on two frequencies, with equal amplitude imposed on the first two normal modes. For Neumann boundary conditions ($\kappa=0$) these solutions are found down to amplitude $\varepsilon=0.001$, and their mode spectra show an exponential fall-off whose decay rate approaches a nonzero constant as $\varepsilon\to 0$, so the small-amplitude limit does not resemble two-mode data. A numerical evolution of a representative solution with $\varepsilon=0.06$ shows periodic $\varphi\varphi^*$ at the boundary and no horizon formation up to $t\approx\varepsilon^{-3}$, which the authors take as evidence of nonlinear stability. This contradicts the expectation that, in a fully resonant system, stable solutions must be single-mode dominated.
Load-bearing premise
The load-bearing assumption is that the numerical family of Neumann double oscillators continues to exist as the amplitude $\varepsilon$ tends to zero; the smallest computed case is $\varepsilon=0.001$, so if the branch stops at some positive amplitude, the claim weakens from 'arbitrarily close to AdS' to existence at finite small amplitudes.
Editorial extensions
If this is right
- For non-resonant Robin spectra, islands of stability include strongly multi-mode data of arbitrarily small energy, supporting the idea that breaking resonance restores stability.
- As the boundary condition approaches the fully resonant Dirichlet case, the maximum energy of two-mode, equal-amplitude double oscillators tends to zero, consistent with such data collapsing.
- In the Neumann case, despite a fully resonant spectrum, strongly multi-mode double oscillators exist arbitrarily close to AdS and are not single-mode dominated.
- These Neumann double oscillators are not described by two-mode perturbation theory: their higher-mode amplitudes decay exponentially with a rate that stays nonzero as $\varepsilon\to 0$, so a perturbative description would require infinitely many modes at leading order.
- Numerical evolution shows a Neumann double oscillator with $\varepsilon=0.06$ remains periodic and non-collapsing up to $t\approx\varepsilon^{-3}$, evidence that the solutions are nonlinearly stable at least on that timescale.
Reading between the lines
- The Neumann result suggests that it is the structure of the secular terms, not merely the presence of a resonance, that decides whether multi-mode data collapse; a resonant spectrum may be necessary for instability but is not sufficient.
- If these bulk solutions have a CFT dual, a double-trace-deformed boundary theory would admit long-lived, non-thermal states at arbitrarily small energy, which is relevant for how double-trace deformations affect thermalization.
- A natural testable extension is to scan other scalar masses and boundary-condition families; the flat-space Dirichlet-box analogue considered in the paper does not reproduce the Neumann plateau, suggesting the AdS boundary structure is essential.
- The observation that extra secular terms shrink when more modes are added at $O(\varepsilon)$ hints at a resummation or non-perturbative description in which the Neumann double oscillator is a coherent multi-mode state; constructing such a description would make the claim testable beyond numerics.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper constructs quasi-periodic 'double oscillator' solutions for a massive complex scalar field (m^2 = -2) in spherically symmetric global AdS, with Robin boundary conditions parameterized by κ that interpolate between resonant Dirichlet (κ = 1) and resonant Neumann (κ = 0) boundary conditions. The central findings are: (i) two-mode equal-amplitude double oscillators exist for generic non-resonant Robin boundary conditions, and their maximum energy vanishes as κ → 1, consistent with the expected instability of two-mode data in the Dirichlet case; and (ii) surprisingly, at the Neumann endpoint κ = 0 the family does not vanish, and the solutions remain strongly multi-mode as the amplitude ε is decreased down to ε = 0.001, despite the fully resonant spectrum. The authors support the existence claim with a Newton-Raphson pseudo-spectral solver, energy and constraint checks at the 10^{-10} level, and a numerical time evolution at one amplitude (ε = 0.06) that shows no collapse up to t ≈ 4630. They also give a perturbative discussion suggesting that the Neumann solutions require seeding an infinite number of modes at leading order.
Significance. If the central claim is correct, the paper provides the first example in AdS of non-collapsing, strongly multi-mode initial data in a system with a fully resonant spectrum, contradicting the emerging picture that islands of stability in resonant cases are single-mode dominated. It also gives a quantitative map of how the space of double oscillators depends on the boundary-condition parameter κ, connecting non-resonant stability to the existence of multi-mode islands. The numerical evidence is substantial: the solver uses a consistent spectral discretization, the Hamiltonian constraint and energy are checked to 10^{-10}, and the paper demonstrates robustness of the existence curve to grid size and parametrization. The main weakness is that the 'arbitrarily close to AdS' statement rests on extrapolation of a trend ending at ε = 0.001, not on a demonstrated limit or a convergence study in ε.
major comments (3)
- [Double Oscillators, Figures 4 and 5] The claim in the abstract and discussion that Neumann double oscillators exist 'arbitrarily close to AdS' is inferred from the plateau of the exponential decay rate in Fig. 5 and the spectrum in Fig. 4 down to ε = 0.001. The paper does not report solutions at smaller ε, does not state residual tolerances for the Newton-Raphson solver, and does not provide a convergence study in grid resolution for the small-ε branch. Since the linearized problem at ε = 0 has an infinite-dimensional kernel (every normal-mode combination is a solution), the numerical branch could terminate or become dominated by numerical noise before reaching arbitrarily small amplitude. Please provide either a convergent sequence of solutions with ε → 0 (with stated residuals and grid sizes) or explicitly weaken the central claim to existence for small but finite ε.
- [Figures 4 and 5, normal-mode decay-rate fit] The statement that the Neumann solutions are 'not single-mode dominated' relies on the exponential decay rate of the normal-mode amplitudes |a_n|, but the n-range used for the linear-regression fit in Fig. 5 is not stated. The text itself notes that at very high n there is a power-law tail because the normal-mode basis is even about the boundary while the nonlinear solutions are only Neumann. If the fitting window includes part of this tail, the 'plateau' in the Neumann decay rate could be an artifact of the fit rather than a physical property of the solution. Please specify the fitted n-interval, show the decay rate as a function of the fit window, and demonstrate that the plateau is robust to the choice of window.
- [Numerical evolution] The nonlinear-stability evidence is presented for a single evolution at ε = 0.06 with energy E = 0.24, evolved to t ≈ ε^{-3} ≈ 4630. The central claim of the paper concerns the limit ε → 0, so the absence of collapse at one moderate amplitude does not establish that the small-ε family is stable on the instability timescale. Moreover, the Discussion states the solutions are 'stable (until at least t ∼ 1/ε^2)', which is inconsistent with the actual evolution time reported (ε^{-3}); please clarify the intended stability timescale and, if the stability claim is to be part of the abstract-level conclusions, provide evolutions at smaller ε or soften the wording to an existence statement with a single finite-amplitude stability check.
minor comments (5)
- [Introduction and Figure 1] There are several typos: 'investigate the affect' should be 'investigate the effect', and the Figure 1 caption has 'mulit-oscillators' instead of 'multi-oscillators'.
- [Double Oscillators] The text says the functions are 'periodic in time with period ω2', but since the Fourier expansion uses cos(k ω2 t), the quantity ω2 is an angular frequency and the period is 2π/ω2. Please correct this wording.
- [Figure 2] The y-axis label in Figure 2 appears garbled: 'ϵmax( e3 e1 )=0.1' should be replaced by a clear expression such as the maximum ε for which |(f_3^{(3)}, ê_3)/(f_3^{(1)}, ê_1)| < 0.1.
- [Perturbative analysis] The phrase 'positive definite frequencies' should be 'positive frequencies'; positive definiteness is a property of operators or bilinear forms, not of frequencies.
- [Numerical methods and references] The paper would benefit from stating the grid sizes, number of Fourier modes, and residual/error tolerances used for the boundary-value problem, and from completing references [1], [2], and [49], which currently lack full bibliographic data.
Circularity Check
No significant circularity: the double-oscillator solutions are constructed by solving the PDE system, and the reported mode spectra, decay rates, and existence curves are computed outputs rather than fitted inputs.
full rationale
The paper's central claims are numerical constructions. The setup in Eqs. (1)-(6) is a standard ansatz for spherically symmetric scalar-field metrics in AdS, and double-oscillators are defined by demanding quasi-periodicity in Eq. (12). The solutions are then obtained by solving the equations of motion as a boundary value problem with a Newton-Raphson method, not by fitting free parameters to the quantities later reported. The amplitude epsilon parametrizes the family through the projections in Eq. (13), but this condition fixes only two mode amplitudes; the higher-mode spectrum, the maximum-energy existence curve in Fig. 1, the mode-amplitude falloff in Fig. 4, and the decay-rate behavior in Fig. 5 are all computed diagnostics. The abstract's 'arbitrarily close to AdS' statement is an extrapolation from computations down to epsilon = 0.001, and the discussion explicitly concedes that the perturbative origin of the Neumann solutions is not yet understood and is left for future work; this is a numerical-continuation and convergence concern, not a circularity. Self-citations to [42,49] supply the multi-oscillator construction method and prior context, but the Neumann boundary-condition result is independently obtained by the boundary value solver and is corroborated by a separate time evolution. No equation in the paper reduces by construction to a claimed prediction, and no load-bearing uniqueness theorem or fitted parameter is renamed as a result.
Assumptions & free parameters
assumptions (5)
- standard math The linearized Klein-Gordon operator L in (7) has a complete set of normal modes with inner product (8), used for projections in (13)-(14).
- domain assumption The symmetry-reduced ansatz (1a)-(1b) and the quasi-periodic Fourier expansion (12) capture all solutions relevant to the island-of-stability question.
- domain assumption The mass m^2 = -2 lies between the Breitenlohner-Freedman and unitary bounds, giving normalizable modes for the Robin family (4) and a double-trace dual.
- domain assumption The Newton-Raphson solver with pseudo-spectral discretization converges to exact smooth solutions, and the reported residuals (constraint and energy conservation at 10^-10) are small enough.
- domain assumption The Hamiltonian constraint, not solved directly, is adequately enforced via a consistency check in the construction and evolution.
Cite this review
Pith. "Pith review of New islands of stability with double-trace deformations." pith.science (2026). https://pith.science/paper/APQ3F4JA
@misc{pith2026190802296,
author = {Pith},
title = {Pith review of: New islands of stability with double-trace deformations},
year = {2026},
howpublished = {\url{https://pith.science/paper/APQ3F4JA}},
note = {Machine review of arXiv:1908.02296}
}
read the original abstract
We assess the role of a resonant spectrum in the AdS instability, and quantify the extent to which breaking the resonant spectrum of AdS can restore stability. Specifically, we study non-collapsing `multi-oscillator' solutions in AdS under various boundary conditions that allow for both resonant and non-resonant spectra. We find non-collapsing two mode, equal amplitude solutions in the non-resonant Robin case, and that these solutions vanish in the fully resonant Dirichlet case. This is consistent with non-resonant stability, and with the idea that stable solutions in the Dirichlet case are all single-mode dominated. Surprisingly, when the boundary condition is Neumann, we find non-collapsing solutions arbitrarily close to AdS that are not single-mode dominated, despite the spectrum being fully resonant.
Figures
Figures from the paper (2 more)
Reference graph
Works this paper leans on
-
[1]
G. H. M. Dafermos (2006)
work page 2006
- [3]
- [4]
- [5]
-
[6]
O. J. C. Dias, G. T. Horowitz, D. Marolf, and J. E. San- tos, Class. Quant. Grav. 29, 235019 (2012), 1208.5772
arXiv 2012
-
[7]
M. Maliborski and A. Rostworowski, Phys. Rev. Lett. 111, 051102 (2013), 1303.3186
arXiv 2013
- [8]
-
[9]
V. Balasubramanian, A. Buchel, S. R. Green, L. Lehner, and S. L. Liebling, Phys. Rev. Lett. 113, 071601 (2014), 1403.6471
arXiv 2014
Show all 55 references
- [10]
-
[11]
Balasubramanian, A
V. Balasubramanian, A. Buchel, S. R. Green, L. Lehner, and S. L. Liebling, Phys. Rev. Lett. 115, 049102 (2015), 1506.07907
2015 arXiv
-
[12]
Dimitrakopoulos and I.-S
F. Dimitrakopoulos and I.-S. Yang, Phys. Rev. D92, 083013 (2015), 1507.02684
2015 arXiv
-
[13]
S. R. Green, A. Maillard, L. Lehner, and S. L. Liebling, Phys. Rev. D92, 084001 (2015), 1507.08261
2015 arXiv
-
[14]
J. M. Maldacena, Int. J. Theor. Phys. 38, 1113 (1999), [Adv. Theor. Math. Phys.2,231(1998)], hep-th/9711200
1999 arXiv
-
[15]
S. S. Gubser, I. R. Klebanov, and A. M. Polyakov, Phys. Lett. B428, 105 (1998), hep-th/9802109
1998 arXiv
-
[16]
Witten, Adv
E. Witten, Adv. Theor. Math. Phys. 2, 253 (1998), hep- th/9802150
1998
-
[17]
O. J. C. Dias, G. T. Horowitz, and J. E. Santos, Class. Quant. Grav. 29, 194002 (2012), 1109.1825
2012 arXiv
-
[18]
Buchel, L
A. Buchel, L. Lehner, and S. L. Liebling, Phys. Rev.D86, 123011 (2012), 1210.0890
2012 arXiv
- [19]
-
[20]
Baier, S
R. Baier, S. A. Stricker, and O. Taanila, Class. Quant. Grav. 31, 025007 (2014), 1309.1629
2014 arXiv
- [21]
-
[22]
P. Basu, D. Das, S. R. Das, and T. Nishioka, JHEP 03, 146 (2013), 1211.7076
2013 arXiv
- [23]
- [24]
-
[25]
Maliborski and A
M. Maliborski and A. Rostworowski, Phys. Rev. D89, 124006 (2014), 1403.5434
2014 arXiv
-
[26]
da Silva, E
E. da Silva, E. Lopez, J. Mas, and A. Serantes, JHEP 04, 038 (2015), 1412.6002
2015 arXiv
- [27]
-
[28]
F. V. Dimitrakopoulos, B. Freivogel, M. Lippert, and I.- S. Yang, JHEP 08, 077 (2015), 1410.1880
2015 arXiv
-
[29]
G. T. Horowitz and J. E. Santos, Surveys Diff. Geom. 20, 321 (2015), 1408.5906
2015 arXiv
-
[30]
P. Bizo, M. Maliborski, and A. Rostworowski, Phys. Rev. Lett. 115, 081103 (2015), 1506.03519
2015 arXiv
-
[31]
P. Basu, C. Krishnan, and P. N. Bala Subramanian, 7 Phys. Lett. B746, 261 (2015), 1501.07499
2015 arXiv
- [32]
- [33]
- [34]
- [35]
-
[36]
D. S. Menon and V. Suneeta, Phys. Rev. D93, 024044 (2016), 1509.00232
2016 arXiv
-
[37]
Freivogel and I.-S
B. Freivogel and I.-S. Yang, Phys. Rev. D93, 103007 (2016), 1512.04383
2016 arXiv
- [38]
-
[39]
Martinon, G
G. Martinon, G. Fodor, P. Grandclment, and P. Forgcs, Class. Quant. Grav. 34, 125012 (2017), 1701.09100
2017 arXiv
-
[40]
O. J. C. Dias and J. E. Santos, Class. Quant. Grav. 35, 185006 (2018), 1705.03065
2018 arXiv
-
[41]
M. W. Choptuik, O. J. C. Dias, J. E. Santos, and B. Way, Phys. Rev. Lett. 119, 191104 (2017), 1706.06101
2017 arXiv
-
[42]
Choptuik, J
M. Choptuik, J. E. Santos, and B. Way, Phys. Rev. Lett. 121, 021103 (2018), 1803.02830
2018 arXiv
- [43]
- [44]
- [45]
- [46]
-
[47]
Ponglertsakul, E
S. Ponglertsakul, E. Winstanley, and S. R. Dolan, Phys. Rev. D94, 024031 (2016), 1604.01132
2016 arXiv
- [48]
- [49]
-
[50]
Buchel, S
A. Buchel, S. R. Green, L. Lehner, and S. L. Liebling, Phys. Rev. D91, 064026 (2015), 1412.4761
2015 arXiv
- [51]
-
[52]
Biasi, P
A. Biasi, P. Bizon, and O. Evnin, Commun. Math. Phys. 369, 433 (2019), 1805.03634
2019 arXiv
- [53]
-
[54]
O. J. C. Dias, G. T. Horowitz, and J. E. Santos, JHEP 07, 115 (2011), 1105.4167
2011 arXiv
-
[55]
S. L. Liebling and C. Palenzuela, Living Rev. Rel. 15, 6 (2012), [Living Rev. Rel.20,no.1,5(2017)], 1202.5809
2012 arXiv
-
[56]
Fodor, P
G. Fodor, P. Forgcs, and P. Grandclment, Phys. Rev. D92, 025036 (2015), 1503.07746
2015 arXiv
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.