Bekenstein Bound

An information ceiling you cannot cross

Take any physical system. Wrap a sphere of radius R around it. Measure its total mass-energy E (rest energy plus kinetic, plus binding, plus everything else relativity counts). Then the system's thermodynamic entropy is bounded above by

S ≤ 2π · k_B · R · E / (ℏ · c)

The bound was proposed by Jacob Bekenstein in 1981, drawing on his earlier (1972) argument that black holes must carry entropy proportional to their horizon area. It is universal — the right-hand side knows nothing about the system's composition, only its size and its energy. Drop in any constants you like; the dimensions work out because k_B has units of entropy and ℏ·c has units of energy × length. Equivalently, in dimensionless natural units (k_B = ℏ = c = 1), the number of distinguishable quantum microstates of any system is bounded by exp(2π R E).

This is the most stringent universal bound on information storage that physics has produced. Whatever your favourite computational substrate — silicon, DNA, neurons, plasma, dark-matter axions — once you specify R and E it cannot encode more bits than the Bekenstein number 2π R E / (ℏ c ln 2). The bound is saturated only by one thing: a black hole.

Bekenstein's 1981 derivation

The bound emerged from a thought experiment. Take a small box containing some matter with entropy S, mass m, and characteristic size R, and quasi-statically lower it on a string toward the horizon of a large Schwarzschild black hole. The work extracted by lowering can be reinvested. But once the box reaches the horizon and is dropped in, the black hole's area increases by some δA, which corresponds to an entropy increase δS_BH = δA / (4 ℓ_P²). The generalised second law of thermodynamics, formulated by Bekenstein in 1972, requires that the total entropy never decrease: δS_BH ≥ S.

Working out δA for a quasi-statically lowered box of mass m and proper size R, Bekenstein derived

δA / (4 ℓ_P²) ≥ 2π m c R / ℏ
            ⇒  S ≤ 2π k_B R E / (ℏ c)         with E = m c².

The constant 2π is the tight one — Bekenstein's gedanken process is reversible at the black-hole horizon in the limit of infinitely massive black holes. Different derivations (Casini 2008 from quantum-field-theoretic relative entropy, Bousso 2003 from the covariant entropy bound) reproduce the same 2π. The number is not adjustable.

Why black holes saturate the bound

Plug r = r_s = 2 G M / c² and E = M c² into the right-hand side:

S_max = 2π k_B r_s · M c² / (ℏ c)
       = 2π k_B · 2 G M / c² · M c² / (ℏ c)
       = 4π G M² k_B / (ℏ c)
       = π k_B (r_s² c³) / (ℏ G)
       = k_B · A / (4 ℓ_P²)               with A = 4π r_s² and ℓ_P² = ℏ G / c³.

The last line is the Bekenstein-Hawking entropy of a Schwarzschild black hole — the exact quantity Hawking computed in 1974 from quantum-field theory on a black-hole spacetime. Two completely different routes (thermodynamic bound vs quantum-field theory) land on the same number. This is the cleanest empirical-looking fact in semiclassical gravity: black holes are not just objects with entropy, they are maximum-entropy objects.

Worked example: how much can a brain hold?

Compute the Bekenstein bound for an adult human brain. Use m = 1.5 kg, characteristic radius R = 0.15 m (a 15-cm sphere encloses an average adult brain), and convert to bits (S/(k_B ln 2)).

E   = m c²
    = 1.5 × (3 × 10⁸)² J
    = 1.35 × 10¹⁷ J

S/k = 2π · R · E / (ℏ · c)
    = 2π · 0.15 · 1.35 × 10¹⁷
       / (1.055 × 10⁻³⁴ · 3 × 10⁸)
    = 1.27 × 10¹⁷ / 3.17 × 10⁻²⁶
    = 4.02 × 10⁴² nats.

bits = (S/k) / ln 2
     = 4.02 × 10⁴² / 0.693
     ≈ 5.8 × 10⁴² bits.

Round to S < 10⁴³ bits. Compare with a generous synaptic estimate: the human brain has roughly 10¹¹ neurons and 10¹⁵ synapses, each storing perhaps a few bits of weight information, giving 10¹⁵ to 10¹⁶ bits of biological storage. The Bekenstein bound is therefore between 26 and 28 orders of magnitude above what biology comes near. The gap is not a measurement issue; it is the difference between "stuff arranged as a brain" and "every quantum-gravitational degree of freedom inside the same sphere", the latter of which would require collapsing the matter into a black hole.

The bound applied to other systems

System	R (m)	E (J)	Bekenstein S/k (nats)	S in bits
Hydrogen atom (ground state)	5.3 × 10⁻¹¹	1.5 × 10⁻¹⁰ (m_e c²)	~ 1.6 × 10⁷	~ 2 × 10⁷
Proton	0.85 × 10⁻¹⁵	1.5 × 10⁻¹⁰ (m_p c²)	~ 4 × 10⁴	~ 6 × 10⁴
1 kg book (1 L)	0.06	9 × 10¹⁶	~ 3 × 10⁴¹	~ 4 × 10⁴¹
Human brain	0.15	1.35 × 10¹⁷	~ 4 × 10⁴²	~ 6 × 10⁴²
Earth (uniform density)	6.4 × 10⁶	5.4 × 10⁴¹	~ 6 × 10⁷⁵	~ 9 × 10⁷⁵
Sun (radiation, photosphere R)	7 × 10⁸	1.8 × 10⁴⁷	~ 2 × 10⁸²	~ 3 × 10⁸²
1 M☉ Schwarzschild BH	2950 (r_s)	1.8 × 10⁴⁷	~ 1 × 10⁷⁷	~ 1.5 × 10⁷⁷
Sgr A* (4 × 10⁶ M☉)	1.2 × 10¹⁰ (r_s)	7.2 × 10⁵³	~ 1.7 × 10⁹⁰	~ 2.4 × 10⁹⁰

The Sun's photosphere is ten million times larger than its Schwarzschild radius, so its Bekenstein bound is about 10⁵ times higher than its black-hole entropy would be. Saturate the bound by compressing it; you create a 1 M☉ black hole and reach the maximum number of microstates physically permissible for that energy.

From entropy bound to holographic principle

The most striking observation is that the bound is set by the boundary, not by the volume. A region of given R can hold at most ~ R² Planck-area worth of bits (because the saturated case is a horizon area), not the ~ R³ that volumetric counting would predict for any standard quantum field. This is the seed of the holographic principle: the genuinely fundamental degrees of freedom of any consistent quantum theory of gravity live on the boundary of a region, not in its interior.

Gerard 't Hooft articulated this in 1993, Leonard Susskind made it concrete in 1995, and Juan Maldacena's 1997 AdS/CFT correspondence is the first explicit realisation: type IIB string theory on AdS_5 × S^5 (a 10-dimensional gravitational theory) is exactly equivalent to N=4 super-Yang-Mills theory on the 4-dimensional boundary. The bulk and the boundary share the same Hilbert space, but the bulk degrees of freedom are heavily redundant. AdS/CFT's two decades of computational successes turn the Bekenstein bound from "thermodynamic bookkeeping" into "a tightly tested principle of the actual world", at least in cousin spacetimes to ours.

Variants and generalisations

• Bekenstein-Hawking entropy (1974). S = k A / (4 ℓ_P²) for black-hole horizons. The Bekenstein bound's saturation case.
• 't Hooft holographic bound (1993). The entropy of any region is bounded by its area in Planck units: S ≤ A / (4 ℓ_P²). Energy-independent; weaker than Bekenstein for low-energy systems, but more universal.
• Bousso (covariant) bound, 1999. The entropy passing through a lightsheet of area A is ≤ A / (4 ℓ_P²). Covers cosmological cases (FRW horizons) where the spherical Bekenstein bound is ambiguous. Reduces to the Bekenstein bound for static, weakly gravitating systems.
• Casini bound (2008). A field-theoretic derivation of the Bekenstein bound from the positivity of the relative entropy between the vacuum and excited states of a quantum field, with the modular Hamiltonian of a Rindler wedge playing the role of E. Provides a fully quantum-field-theoretic basis for the bound, independent of black holes.
• Strominger-Vafa microstate count (1996). A direct string-theory computation of the microstates of a class of extremal black holes, reproducing the Bekenstein-Hawking entropy exactly. Strong evidence the bound counts genuine microstates, not just thermodynamic capacity.

Where the Bekenstein bound shows up

• Black-hole thermodynamics. The bound is saturated by every Schwarzschild horizon; together with Hawking's temperature formula and the Bekenstein-Hawking area law, it makes black holes look exactly like ordinary thermodynamic systems.
• Holographic principle. Direct inspiration: areas not volumes govern degrees of freedom. AdS/CFT is the canonical worked example; "bulk reconstruction" research depends on the principle's exactness.
• Quantum gravity sanity checks. Any candidate theory must reproduce the bound and its saturation. Loop quantum gravity, string theory, asymptotic-safety scenarios, causal-set theory all engage with it.
• Computer-science upper limits. Seth Lloyd's 2000 paper "Ultimate physical limits to computation" uses the Bekenstein bound (together with the Margolus-Levitin bound on operations per second) to compute the maximum-possible computer: a 1 kg "ultimate laptop" achieves at most 10⁵¹ operations per second and stores 10⁵¹ bits, both numbers being a few orders of magnitude below the Bekenstein limit for the same R and E.
• Cosmological information. Applied to the de Sitter horizon, the bound gives roughly 10¹²² bits — an estimate of the total information content of the observable universe.

Common pitfalls

• Forgetting the radius is the bounding sphere, not some internal scale. R is the radius of the smallest sphere that completely contains the system, measured externally. Using the system's "size" loosely (radius of gyration, etc.) typically overstates the bound.
• Forgetting that E is total mass-energy. Plug in rest energy + kinetic + thermal + binding; do not just use thermal energy. For most everyday systems mass dominates and S ≈ 2π k m c R / ℏ.
• Conflating volume scaling with the bound. Quantum-field theory naively gives entropy ∝ R³ via UV mode counting; the Bekenstein bound says R³ scaling is illegal in nature because dense modes gravitationally collapse. The reconciliation is the holographic principle.
• Treating saturation as inevitable. A 1 kg book is 26 orders of magnitude under the bound — saturation requires shrinking the system to its Schwarzschild radius (collapsing it into a black hole), not merely "adding more information".
• Reading the bound as a statement about classical entropy. The bound is on the von Neumann entropy of the quantum state. For mixed states it is sharp; for pure states the modular relative entropy formulation (Casini) gives the cleanest derivation.