Mathematician review (Step 2): Verlinde-Erik

Project: E. Verlinde, "On the Origin of Gravity and the Laws of Newton", arXiv:1001.0785v1 (6 Jan 2010; later JHEP 04 (2011) 029). Reviewer role: PhD mathematician, checking derivations, notation, units and hidden assumptions. No source gap: the full text layer exists (29 pages). Tags: [read: p.N] from the paper; [read: page image p.N] from a rendered page (via the deep read); [calc] my own sympy/numpy check (venv, scratch only); [web] URL; [memory] with confidence.

Sources used: facts.json, claims.md, the deep read (Sections 4-5, pp.12-22), the paper text of pp.4-14 and 22-28 which I read myself, and the Investigator's sites.md "Leads for review" and research notes. No file was found in research/files/ (none downloaded), so the Investigator's items are web abstracts only.

0. Notation a reader needs

• F force, m test mass, M enclosed mass, R screen radius, A screen area, a acceleration, Phi Newton potential (energy per unit mass; a = -grad Phi), c speed of light, hbar reduced Planck constant, k_B Boltzmann constant, G the constant that eq. (3.10) defines.
• "Bit": one degree of freedom on the holographic screen. N = number of bits. T = temperature, S = entropy.
• Entropic force: F = T dS/dx. If the entropy of a system changes by dS when a coordinate moves by dx, and the system sits in a heat bath at temperature T, there is an effective force. A stretched rubber band is the standard example.
• Equipartition: each degree of freedom carries (1/2) k_B T of energy.
• Killing vector xi^a, redshift phi, Komar mass: introduced in section 4 below before they are used.

1. What I verified myself, in one table (Step 1 checked, not trusted)

2. New points from my own reading (additions to Step 1)

2.1 Bekenstein-bound violation, quantified. The paper notes only that the screen entropy "appears to violate the Bekenstein bound" [read: p.27]. For a spherical screen of radius R holding mass M, S = k_B pi R^2 c^3/(G hbar) from (6.41) [read: p.27], and the bound is S <= 2 pi k_B M c R/hbar. The ratio is S/S_bound = R/R_s with R_s = 2GM/c^2 [calc]. For Earth's surface, R_s = 8.87 mm, so the ratio is about 7.2e8 [calc]. The violation is therefore a large factor, growing with distance from the horizon, and equal to 1 only at the horizon. The paper's escape routes (rescale hbar by f(Phi_0); non-equilibrium) are not worked out [read: p.27]. This is not a numerical error, but it shows the "bits" cannot be literal thermal degrees of freedom at the real Planck constant away from horizons, which the author concedes ("the role of hbar is subtle and not yet fully understood", [read: p.26]).

2.2 What the Section 3 derivation actually establishes (counting inputs). Inputs: (i) bits proportional to area with coefficient c^3/(G hbar), where G is defined by this coefficient [read: p.8]; (ii) equipartition; (iii) E = M c^2; (iv) a linear entropy gradient with the 2 pi chosen to give F = ma [read: p.8]. Output: inverse-square law with coefficient G. Because hbar cancels and G is defined by (3.10), the calculation cannot predict the value of G. What does follow is the shape: 1/R^2 comes from "number of bits ~ area", and linearity in M and m from additivity. Since the Unruh temperature of g = GM/R^2 equals the equipartition temperature (table above), eq. (3.8) is, once G is defined through (3.10), a consistency identity rather than an independent input. The author says as much: "the laws of Newton have been ingredients in the steps that lead to black hole thermodynamics and the holographic principle. In a sense we have reversed these arguments" [read: p.9]. So the derivation is a re-encoding: Newton's law is exchanged for (area-law bits + equipartition + entropy gradient), and the constants are calibrated to Newton's law.

2.3 Dimension-generalisation footnote. Footnote 4 adds a factor (1/2)(d-2)/(d-3) in d dimensions [read: p.10]. In n spatial dimensions with the convention that Poisson reads laplacian Phi = 4 pi G rho, my own calculation shows no extra factor is needed (F = 4 pi G M m/(kappa A) with kappa = 1) [calc]. The factor therefore reflects the convention used for G in higher-dimensional GR, not extra physics. Confidence medium (I did not trace the paper's convention).

2.4 Conservative-force condition (the mathematics behind Visser's critique). F = T grad S has curl(T grad S) = grad T x grad S. So F is conservative only if grad T and grad S are parallel, i.e. S is locally a function of T. In the paper's setting T is proportional to |grad Phi| on the screen (varies from point to point for non-spherical sources) while the entropy gradient (5.29)/(4.22) is set by delta Phi, so there is no guarantee [calc/reasoning]. Visser's abstract says exactly this class of constraint "gives real and significant problems" for Verlinde's version and not for Jacobson's or Padmanabhan's, and proposes multiple heat baths [web] https://arxiv.org/abs/1108.5240 (abstract read; I did not read the proof). Polymer analogy supports conservative force only at constant T [read: p.5].

2.5 Equipotential-only family and the Poisson step. Gauss's law (4.20) is required only for equipotential screens, a one-parameter family Phi_0. For a single static source that gives one scalar condition per level, whereas the Poisson equation is a pointwise equation. The paper bridges this with the sentence about adding a bit of mass near the screen and pushing it across [read: p.14]. That is an appeal to variations of the mass distribution, which is not the same thing as the equipotential family and is not carried out. The deep read called this gap (a); I agree and state it as a counting argument, not as a constructed counterexample [reasoning, not proven]. Also, Phi appears as "a device to describe the local acceleration" in the definition of T [read: p.14], so the potential is an input to the temperature before the equation for it is derived.

2.6 The two-pi and flat-patch factor of two. The deep read found [calc] that for a flat local patch the entropy-change integral gives half of the 2 pi of (3.6); the full 2 pi needs the closed screen. I did not re-derive this but my Earth check agrees with the global statement: the integral gives R^2/d^2 approaching 1 as the test mass reaches the screen. Marked "reproduced from the deep read, not independently rerun".

2.7 Relativistic part. I confirmed (5.30), (5.31), (5.34), (5.39) as pure algebra from stated inputs [calc]. The input (5.37) left-hand side, a particular combination 2(T_ab - (1/2) T g_ab) n^a xi^b, is "presumably" fixed by comparing both sides [read: p.20]. In Wald's treatment that expression comes from using the Einstein equations. So the matter side of the field equation is assumed. Approximate Killing vectors in small regions do not satisfy (5.36) exactly, and the error terms are not discussed (deep read, agreed). A cosmological-constant term can be added without breaking the argument (Jacobson-style) [memory: medium]; the paper is silent on it. Section 5.3 (5.38)-(5.39) is a generic first-law identity, which the author notes "did not really use the precise form of the Komar integral, or the Einstein equations" [read: p.22]; by the same token it cannot single out GR.

3. Claim-by-claim review (questions a-d and verdict)

Claim 1: Gravity is an entropic force (F = T dS/dx), like polymer elasticity

a. Internally consistent? Yes as algebra [calc]. The polymer analogy equations (2.1)-(2.4) are standard statistical mechanics [read: p.4-5]. Weak points: the force is conservative only for constant T (author admits, p.5); (Visser, section 2.4 above). Equation (3.8) is "to be read" the opposite way from the usual Unruh effect [read: p.8]. Signs of (4.25), (5.27)/(5.30) are loose. b. Consistent with experiment? It makes no new prediction in the Newtonian regime; it reproduces Newton's and Einstein's equations at the level of the paper, so all existing tests are reproduced by construction. Kobakhidze argues ultra-cold neutron bound states in Earth's field disprove entropic gravity [web] https://arxiv.org/abs/1009.5414 (abstract read). Chaichian-Oksanen-Tureanu reply that EG does not necessarily contradict the GRANIT data but would need extra ingredients to explain spontaneous decays of excited states [web, via Investigator note on arXiv:1104.4650; abstract not re-opened by me]. Verlinde's 2026 note, as reported by the Investigator, says the slow system keeps quantum coherence (Born-Oppenheimer regime) [web] https://www.worldscientific.com/doi/10.1142/S0217751X26470020 (Investigator's reading; not re-read by me). Wang & Braunstein: for surfaces away from horizons "the first law fails (with the exception of fully spherically symmetric scenarios), thus undermining the key thermodynamic assumption" [web] https://arxiv.org/abs/2207.04390 (abstract read; the published version is Nat. Commun. 9, 2977). A reader comment recorded by the Investigator says the criticism does not apply to the 2016 volume-law model; I record it as a commenter's view. c. Tested? No direct neutral test of the 2010 paper's entropic-force mechanism on record. Indirect: neutron bound states (argued both ways above, by theorists, not a controlled test of the mechanism); the later 2016 model has astrophysical tests (claim 5 below). d. If true: Newtonian and Einsteinian gravity would be effective descriptions; the equivalence of gravitational and inertial mass would be a consequence of the same T dS/dx mechanism (inertia, claim 2); quantum gravity would not need to quantise a gravitational field. Standard physics would not need to change for existing data, but the question "what are the microscopic degrees of freedom" would become the central one. Verdict: split. Strong form (gravity is an entropic force with a stated mechanism): Never properly tested (and, as presented, not testable at the level of the mathematics given: it reproduces known laws by construction; the 2010 paper itself says "we need a better understanding of the theory to turn this in to a prediction" [read: p.27]). Weak form (gravity admits a thermodynamic reading, with Jacobson-type equation of state): Plausible, untested in the sense of the entropic mechanism; the mathematics of the thermodynamic reading exists independently (Jacobson 1995 [read: p.28 ref 7]). The specific Verlinde version has published mathematical objections (Visser; Wang-Braunstein) that are not refutations of the horizon case.

Claim 2: Inertia is entropic (dS = 2 pi k_B (mc/hbar) dx with Unruh T gives F = ma)

a. Algebra is correct: combining (3.6), (3.7), (3.8) gives F = ma [calc]. But the 2 pi in (3.6) is "picked precisely in such a way that one recovers" F = ma [read: p.8], so the result is calibrated, not derived. The Unruh formula is used with a reversed reading ("the temperature required to cause an acceleration a", [read: p.8]) and the paper elsewhere says "forget about the Unruh law" [read: p.8] while the same form of T is reused in sections 4 and 5. Mild internal tension (a documentation inconsistency, not a contradiction of equations). b. Unruh effect itself has never been observed directly [memory: high]; the postulate (3.5) is a hypothesis about entropy near a screen; no experimental constraint on it. c. Tested: no. d. If true: Newton's second law would be a statistical statement; mass in the equation would be an entropy-per-displacement measure. Q: does it differ from F = ma anywhere measurable? Not at this level. Verdict: Not testable as stated (reproduces F = ma by construction through a chosen constant). Mathematically it is a consistent re-expression (Plausible, untested as an interpretation) rather than a derivation.

Claim 3: Newton's law and Poisson/Gauss from holography plus equipartition, for general matter distributions

a. Section 3 and 4.1 algebra: correct [calc]. Gauss's law (4.20) follows in one line from (4.17)-(4.19) once T is defined from grad Phi; it is thus equivalent to equipartition, not an additional result (see 2.2). The Poisson equation is reached by "requiring" (4.20) for all equipotential screens [read: p.14]; the counting argument in 2.5 shows this is not a proof. Section 4.2 (virtual displacement, (4.22)-(4.25)) uses the Poisson equation and Green's-function methods as the engine and reproduces Newton's forces on test bodies in the cases checked (point mass inside spherical screen), with a sign convention issue [calc]. The proof is described as "straightforward application of electrostatics" and not written out [read: p.16]. Inputs include the Newton potential (as the definition of T), G (defined by bit density), E = Mc^2 and a bits-per-area count with uniform distribution that is declared, not derived, for non-spherical screens ("We again assume that these bits are uniformly distributed", [read: p.14]). b. Consistent with all Newtonian-regime experiments (since it returns Newton's law). The N-bit count implies S = k_B A c^3/(4 G hbar) for equilibrium screens, which violates the Bekenstein bound by a factor R/R_s away from horizons (2.1). Equipartition "in general holds only for free systems" [read: p.10]. Wang & Braunstein: non-spherical screens fail the first law [web] https://arxiv.org/abs/2207.04390. c. Tested: no test beyond reproducing Newton. Visser: general-potential versions run into the conservative-force constraint [web] https://arxiv.org/abs/1108.5240. d. If true: G would be a count of bits per area (G = A c^3/(N hbar)); Newtonian gravity would be a derived statistical result. No numerical value of G would be predicted. Verdict: split. Spherical symmetry / single point-mass case: Refuted by calculation does not apply; the result is internally consistent, but equivalent to Gauss's law (Established as a mathematical identity, not as a derivation of gravity). General-distribution case (Poisson from the equipotential-screen argument): Never properly tested; argued only at heuristic level, with published objections (Visser; Wang-Braunstein). It is not Internally inconsistent: the author flags the gaps himself.

Claim 4: Relativistic version: Komar mass, F = -m e^phi grad phi, and Einstein equations via a Jacobson-style argument

a. (5.26)-(5.31), (5.33)-(5.34), (5.39): correct algebra from the stated inputs [calc]. The argument for the full Einstein equations is a one-paragraph sketch (the author calls it "a sketch", [read: p.20]) that (i) posits the stress-energy combination in (5.37), (ii) uses approximate Killing vectors in small regions without error control, (iii) requires "arbitrary screens" though screens were defined as constant-redshift surfaces. The starting point is already a static Lorentzian spacetime, a Killing vector and the Komar formula from Wald (footnote 5 [read: p.17]). The author states that (5.34) "can be derived by assuming the Einstein equations" and that he comes "from the other side" [read: p.19]; the genuine new content is the identification of the Komar integrand with (1/2) T dN. b. Consistent with GR tests only because it reproduces GR (static case). It makes no independent check. c. Tested: no. d. If true: GR's field equations would be an equation of state, as in Jacobson 1995 [read: p.28 ref 7]; this is a widely discussed idea and does not by itself change experimental physics. Verdict: Plausible, untested as a thermodynamic derivation, but the Einstein-equation step as written is not a proof (assumed input; sketch). For the 2010 area-law screens away from horizons, the first-law assumption is contested by Wang-Braunstein (see claim 1).

Claim 5: Space, geometry and gravity are emergent; string theory / AdS/CFT support; redshift is a consequence of entropy gradients

a. The mathematics here is mostly verbal. The one equation, (5.31), expresses the force balance in microcanonical form [read: p.19]. The author writes that the presentation runs "the other way" from the microscopics and that this is something that "might have appeared somewhat contrived" [read: p.19]. The "redshift is a consequence of entropy gradient, not the other way around" is an interpretive statement; both readings are said to be "equivalent in the relativistic setting" [read: p.19], so there is no measurable difference in the 2010 paper. Section 6.2 on open/closed strings and D-branes as screens is qualitative [read: p.24]. b. Nothing to compare in the 2010 paper. Later, the 2016 model (area plus volume law; "dark gravity" with a_0 = c H_0) is a different mathematical construction with a concrete low-acceleration prediction [web] https://arxiv.org/abs/1611.02269 (Investigator's note). The Investigator's findings, kept here: Brouwer et al. 2017 report that the no-free-parameter EG prediction is in good agreement with KiDS x GAMA lensing in four mass bins, and call it "only a first step" [web] https://arxiv.org/abs/1612.03034 (abstract read by me); Brouwer et al. 2021 (Verlinde is a co-author, an interested party) report agreement with the RAR but also a difference of at least 6 sigma between early- and late-type galaxies that current modified-gravity theories "would be unable to explain" without circumgalactic gas, and that LambdaCDM also fits [web] https://arxiv.org/abs/2106.11677 (Investigator's note, not re-opened by me); Pardo finds EG fits Tully-Fisher and massive spirals but "severely underpredicts" maximum velocities above ~100 km/s for 452 isolated dwarfs [web] https://arxiv.org/abs/1706.00785 (Investigator's note); Nieuwenhuizen 2017 argues Abell 1689 needs eV neutrinos [web, via Wikipedia per Investigator; primary not opened]. These bear on the 2016 programme, not on the mathematics of the 2010 paper; I flag that clearly so the 2010 paper is not credited or discredited with them. c. Tested: the 2016 successor has been tested by others, with mixed outcomes as above. The 2010 formulation: no test. d. If true: spacetime would be derived from entanglement/information; this is a research direction already accepted as possible in the AdS/CFT community [memory: high] (in that setting, gravity in the bulk is dual to a non-gravitational boundary theory; the 2010 paper's claim that this implies gravity is entropic in our universe is an extrapolation). Verdict: split. The AdS/CFT-style emergence of a gravitational description in a dual non-gravitational theory: Established as a mathematical correspondence in that setting [memory: high]; extension to our universe: Never properly tested. "Redshift is a consequence of entropy gradient": Not testable as stated (the paper says both views are equivalent). 2016 successor claims: out of the paper's scope; some tests exist, mixed.

Claim 6: Gravitational force fluctuations, larger for weak fields between small bodies

a. Not derived; one paragraph, with the explicit statement "we need a better understanding of the theory to turn this in to a prediction" [read: p.27]. There is no formula to check. The size of fluctuations depends on an "effective temperature" that "may not be universal" and on "the effective value of hbar" [read: p.27], and the same page says hbar can be rescaled by a function of Phi_0 without changing the force. A fluctuation size that depends on the freely rescalable hbar is not determined by the stated mathematics. b. No experiment cited. Later work by the author makes it quantitative for interferometers (Verlinde-Zurek 2019/2021, 2020, 2022; GQuEST design paper, Vermeulen et al. PRX 15, 011034 (2025)); no GQuEST measurement result was found by the Investigator [web] https://arxiv.org/abs/2404.07524. Carney-Karydas-Sivaramakrishnan estimate Delta L of order 1e-35 m from standard graviton EFT [web] https://arxiv.org/abs/2409.03894 (Investigator's note; I did not open), so a positive signal would indicate EFT breakdown. These connect to the geontropic model and are not shown to follow from the 2010 paper's equations. c. Tested: the proposal is being tested at the design stage; no result on record in what I found. d. If true: a detectable noise floor in interferometers; breakdown of effective field theory for gravitons at low energy. Verdict: Not testable as stated in the 2010 paper; the later quantitative form is Never properly tested (a test is being built). Whether these two are the same claim needs the author's own papers; I did not check.

4. Investigator's findings I am keeping in the record (relevance to the mathematics)

Inventor's own work: Verlinde 2026 (IJMPA 41(26) 2647002) says an entropic force "acts without increasing the entropy" and reframes it as an adiabatic reaction force; this speaks directly to the sign and constant-S issues in (4.25), (5.27)/(5.30), (5.38) [web per Investigator]. Verlinde 2016 and Verlinde-Zurek 2019-2022, Strings 2011 talk (not read). JHEP 2011 version differs possibly from the arXiv v1 project copy; I did not compare (unknown). Supporting: Brouwer 2017 and 2021, Chaichian et al. 2011, GQuEST, Pardo (partial). Related work: Jacobson 1995, Padmanabhan 2010 (the closest mathematical precursors; the Einstein-equation step in Verlinde's paper borrows from Jacobson), Bianconi 2025, Parker-Jeynes, the 2010 follow-ups, Visser 2011, Wang-Braunstein 2018, Kobakhidze 2011, Pardo 2020, Nieuwenhuizen 2017, Carney et al. 2026. Historical record: Volkskrant 12 Dec 2009, arXiv 6 Jan 2010, press coverage, Spinoza Prize 2011, a continuing programme in mainstream journals; the record does not go silent, so no absence-of-evidence anomaly arises. Two sources (Nieuwenhuizen via Wikipedia; Strominger quote via Wikipedia) were not opened by the Investigator; I repeat them as second-hand.

5. Corrections to Step 1

1. No substantive error found in the deep read's equations; its numbers reproduced ([calc] table above). Confirmed typo in (4.17).
2. claims.md says Sections 1-3 and 6 are digest-only. I read pp.4-11 and 22-28 in the extracted text, so (3.5)-(3.13), (6.40)-(6.41) and the fluctuation paragraph are now checked against the text itself, not only the digest.
3. facts.json/claims.md present claim 3 as "follows from holography". Following the paper's own p.9 statement, "derives" overstates: the derivation is a reparametrisation calibrated on Newton's law (2.2). The wording of claim 3 should say "is re-expressed as", with the author's reversal disclosure kept. This is a matter of emphasis, not a factual error.
4. The sign claim in the deep read ("sign-convention looseness") stands, and I add that Verlinde's 2026 note, per the Investigator, addresses it conceptually. Not independently verified.
5. The "Bekenstein bound" caveat in claims.md should carry the number: factor R/R_s (about 7e8 for Earth's surface) (2.1).
6. Investigator's leads: attribution OK for the four I opened (Visser 1108.5240, Kobakhidze 1009.5414, Brouwer 1612.03034, Wang-Braunstein, arXiv 2207.04390, whose published version is Nat. Commun.). Note the Investigator's arxiv note lists 2207.04390 under "not reached" but the Nature note covers the same paper; no conflict.

6. Verdict summary

1. Gravity as entropic force: strong form Never properly tested (not testable beyond reproducing known laws); weak/thermodynamic form Plausible, untested; contested by Visser and Wang-Braunstein.
2. Inertia entropic: Not testable as stated (calibrated by chosen 2 pi); consistent algebra.
3. Newton from holography + equipartition: single-source case algebra is correct and equivalent to Gauss's law; general-distribution case Never properly tested, step to Poisson is a heuristic.
4. Relativistic and Einstein equations: Plausible, untested; the Einstein step is an acknowledged sketch with assumed input.
5. Emergent spacetime: AdS/CFT emergence Established in its own setting [memory: high], extension to our universe Never properly tested; "redshift from entropy gradient" Not testable as stated; 2016 successor has mixed tests.
6. Force fluctuations: Not testable as stated in 2010; later quantitative form Never properly tested (experiment under construction).