Notes on de Sitter space

Abstract: We discuss classical and quantum features of spacetimes with a positive cosmological constant. After introducing basic aspects of de Sitter cosmology and the classical geometry of de Sitter space, we consider de Sitter space both as a rigid background, on which quantum fields propagate, and as a fluctuating spacetime in its own right. We review the wavefunction of an expanding universe, black holes in de Sitter space, and the de Sitter entropy, and compare the cosmological horizon with the black hole horizon. The final three sections are dedicated to concrete models of de Sitter space in two, three and four spacetime dimensions.

• The paper presents a detailed review of de Sitter space, integrating differential quantum fields, gravity, and black-hole features with a common framework.
• The review highlights distinct characteristics of de Sitter geometry, such as compact spatial slices, future and past boundaries, and the absence of globally timelike Killing vectors in its quantization.
• A massless scalar field in the static patch of dS_4 exhibits a scale-invariant spectrum and provides a singular example of the scalar field in the asymptotic conformal theory

Scope and organizing perspective

“Notes on de Sitter space” (2609.19454) is an extensive review of classical de Sitter geometry, quantum fields on fixed de Sitter backgrounds, semiclassical quantum gravity, horizon thermodynamics, black holes, and explicit lower- and higher-dimensional models. Its central organizing distinction is between rigid de Sitter quantum field theory and quantum de Sitter gravity. In the former, the metric is fixed and matter fields furnish unitary irreducible representations of $\mathrm{SO}(1,d+1)$. In the latter, the geometry itself fluctuates, the Hamiltonian becomes a constraint, and the standard Hilbert-space, entropy, and thermodynamic interpretations require substantial modification.

The notes are deliberately encyclopedic rather than centered on a single new theorem. Their contribution is synthetic: they place several technically distinct descriptions—global and static-patch quantization, late-time conformal data, Euclidean sphere partition functions, algebraic observables, quasinormal modes, low-dimensional path integrals, matrix models, and higher-spin holography—within a common framework. The recurring issue is that de Sitter space has compact spatial slices, spacelike future and past boundaries, cosmological horizons, and no globally defined timelike Killing vector. These properties obstruct direct transplantation of asymptotically flat or AdS methods.

Classical cosmology and de Sitter geometry

The cosmological discussion begins with the FLRW metric and derives the continuity, Hamiltonian, and acceleration equations. The treatment emphasizes the different dilution laws of matter components,

$\rho_i \propto a^{-3(1+w_i)},$

and the special role of vacuum energy, whose density remains constant as the scale factor grows. Consequently, even a small positive cosmological constant eventually dominates the expansion. The notes quote representative present-day values $\Omega_\Lambda \approx 0.685$, $\Omega_{\rm dust}\approx 0.315$, and $\Omega_{\rm rad}\approx 9\times 10^{-5}$, while identifying the observed cosmological constant as approximately $10^{-122}$ in Planck units. The latter provides the connection to the cosmological constant problem: quantum vacuum contributions are parametrically much larger than the observed value, but their cancellation mechanism is not addressed by the geometric analysis.

Pure $\mathrm{dS}_{d+1}$ is represented as the hyperboloid

$-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$

in $\mathbb{R}^{1,d+1}$. This embedding makes the isometry group $\mathrm{SO}(1,d+1)$ manifest and establishes the relation

$\rho_i \propto a^{-3(1+w_i)},$0

Global coordinates exhibit a closed universe with spatial slices $\rho_i \propto a^{-3(1+w_i)},$1, contracting from $\rho_i \propto a^{-3(1+w_i)},$2 to a minimal sphere and re-expanding toward $\rho_i \propto a^{-3(1+w_i)},$3. The static patch covers only the causal domain of one observer and has metric

$\rho_i \propto a^{-3(1+w_i)},$4

The surface $\rho_i \propto a^{-3(1+w_i)},$5 is a cosmological event horizon, not a coordinate artifact relative to the static observer.

A technically important point is that no de Sitter Killing vector is everywhere timelike. Therefore global de Sitter lacks a preferred Hamiltonian and a globally meaningful positive-energy decomposition. The static-patch generator is timelike only within the observer’s causal diamond. This distinction later explains why static-patch quasinormal modes and thermal correlation functions coexist with the absence of global energy eigenstates.

The notes also formulate asymptotically de Sitter geometries in Fefferman–Graham form. In four spacetime dimensions, the asymptotic data consist of a conformal class $\rho_i \propto a^{-3(1+w_i)},$6 at $\rho_i \propto a^{-3(1+w_i)},$7 and a transverse-traceless tensor $\rho_i \propto a^{-3(1+w_i)},$8. After quotienting by boundary diffeomorphisms and Weyl transformations, $\rho_i \propto a^{-3(1+w_i)},$9 contains two local gravitational degrees of freedom, while $\Omega_\Lambda \approx 0.685$0 supplies their conjugate momentum. This formulation connects the asymptotic initial-value problem to Friedrich’s theorem and makes precise why late-time de Sitter data have the structure of Euclidean conformal data, despite originating from Lorentzian evolution.

Quantum fields and de Sitter representation theory

The fixed-background quantum theory is organized by unitary irreducible representations of $\Omega_\Lambda \approx 0.685$1. For a scalar field, the quadratic Casimir relates the representation weight to the mass through

$\Omega_\Lambda \approx 0.685$2

The principal series has $\Omega_\Lambda \approx 0.685$3, the complementary series has real $\Omega_\Lambda \approx 0.685$4 in an appropriate interval, and exceptional or discrete representations contain massless and partially massless fields. In $\Omega_\Lambda \approx 0.685$5, the classification can be derived explicitly from $\Omega_\Lambda \approx 0.685$6, including principal, complementary, and highest- and lowest-weight discrete series.

For spinning fields, the Casimir is

$\Omega_\Lambda \approx 0.685$7

and partially massless representations occur at discrete masses

$\Omega_\Lambda \approx 0.685$8

The depth $\Omega_\Lambda \approx 0.685$9 controls the derivative order of the gauge symmetry. The Higuchi bound,

$\Omega_{\rm dust}\approx 0.315$0

is interpreted as the condition preventing negative-norm Stückelberg modes. At special discrete masses, enhanced gauge symmetry removes the would-be ghost. The graviton is identified as the highest-depth partially massless spin-two field, while higher-spin de Sitter theories contain the corresponding tower of representations.

The paper makes a useful conceptual comparison with AdS. In AdS, the Hamiltonian belongs to a compact $\Omega_{\rm dust}\approx 0.315$1 subgroup and has a spectrum bounded below. In de Sitter, the analogous generator belongs to noncompact $\Omega_{\rm dust}\approx 0.315$2 and is not positive definite. The de Sitter future boundary nevertheless carries an action of the Euclidean conformal group, because $\Omega_{\rm dust}\approx 0.315$3 is simultaneously the de Sitter isometry group and the conformal group of $\Omega_{\rm dust}\approx 0.315$4.

The Harish–Chandra character provides the bridge between group theory and static-patch spectroscopy. Formally,

$\Omega_{\rm dust}\approx 0.315$5

but for noncompact groups this trace is distributional. Its expansion in exponentials encodes the degeneracies of static-patch quasinormal modes. For a conformally coupled scalar in $\Omega_{\rm dust}\approx 0.315$6, the character expands as

$\Omega_{\rm dust}\approx 0.315$7

The coefficients $\Omega_{\rm dust}\approx 0.315$8, $\Omega_{\rm dust}\approx 0.315$9, and $\Omega_{\rm rad}\approx 9\times 10^{-5}$0 arise from the quasinormal-mode degeneracies at frequencies $\Omega_{\rm rad}\approx 9\times 10^{-5}$1, $\Omega_{\rm rad}\approx 9\times 10^{-5}$2, and $\Omega_{\rm rad}\approx 9\times 10^{-5}$3, respectively. This is one of the clearest quantitative correspondences developed in the notes: the character is not merely an abstract representation-theoretic object, but a generating function for dissipative static-patch excitations.

Late-time operators, correlators, and wavefunctions

The notes derive late-time conformal operators directly from bulk field equations. A scalar field has two asymptotic branches,

$\Omega_{\rm rad}\approx 9\times 10^{-5}$4

where $\Omega_{\rm rad}\approx 9\times 10^{-5}$5 in the planar patch. The two coefficients are canonically conjugate in the asymptotic phase space. Their commutator is fixed by the bulk canonical commutation relations: $\Omega_{\rm rad}\approx 9\times 10^{-5}$6 The reality properties differ between the principal and complementary series. For principal-series fields, the two branches are related by Hermitian conjugation; for complementary-series fields, each branch can be separately Hermitian while remaining canonically conjugate.

The massless scalar in $\Omega_{\rm rad}\approx 9\times 10^{-5}$7 provides a concrete result. Its late-time wavefunction yields

$\Omega_{\rm rad}\approx 9\times 10^{-5}$8

which is scale invariant. The paper correctly stresses the associated infrared qualification: the strict $\Omega_{\rm rad}\approx 9\times 10^{-5}$9 mode has vanishing Gaussian suppression, so the massless de Sitter wavefunctional is not normalizable without an infrared prescription. The scale-invariant spectrum is therefore a property of the nonzero-momentum modes, not a complete statement about the global wavefunctional.

The Bunch–Davies state is defined by positive-frequency behavior in the far past along an $10^{-122}$0-deformed contour. For free fields, the wavefunctional is obtained by evaluating the on-shell action on the saddle satisfying this condition. In the massless $10^{-122}$1 example, the resulting Gaussian kernel is

$10^{-122}$2

and the normalized probability distribution produces the two-point function above. A conformally coupled scalar instead has variance

$10^{-122}$3

demonstrating that scale invariance is not generic but depends on the late-time representation.

For gravity, the corresponding object is the Hartle–Hawking wavefunction, defined by a Euclidean path integral over regular geometries that cap off smoothly. The Wheeler–DeWitt equation replaces an ordinary Schrödinger evolution equation. In the large-volume expansion, the semiclassical phase contains local terms proportional to the spatial volume and integrated scalar curvature, while the nonlocal functional $10^{-122}$4 controls the modulus of the wavefunction. The Hamilton–Jacobi constraint implies that $10^{-122}$5 is Weyl invariant at leading late times, motivating an inner product over metrics modulo $10^{-122}$6. The notes explicitly acknowledge that this does not uniquely determine the measure or resolve the contour problem.

Euclidean gravity, entropy, and the conformal mode

The Gibbons–Hawking proposal is formulated as

$10^{-122}$7

Unlike the black-hole relation $10^{-122}$8, de Sitter has no asymptotic region supporting an independently defined ADM energy. The Euclidean continuation of de Sitter is a sphere, and the absence of an energy term leaves the partition function itself to represent the entropy.

At tree level,

$10^{-122}$9

In four dimensions,

$\mathrm{dS}_{d+1}$0

and for the observed cosmological radius this is of order $\mathrm{dS}_{d+1}$1. The paper emphasizes that this enormous number should not automatically be interpreted as the logarithm of an ordinary finite-dimensional Hilbert space. In continuum QFT, the static-patch algebra is type III, so it has no trace and no density matrices in the usual sense.

The one-loop analysis is one of the most technically developed parts of the review. Functional determinants on $\mathrm{dS}_{d+1}$2 are reorganized using heat kernels, Hubbard–Stratonovich transforms, Hurwitz zeta functions, and Harish–Chandra characters. The one-loop result separates into bulk and edge contributions,

$\mathrm{dS}_{d+1}$3

The bulk character counts propagating degrees of freedom, whereas the edge term has a codimension-two ultraviolet divergence and is associated with gauge and horizon-localized data. This obstructs a literal interpretation of the full sphere partition function as a trace over bulk quasinormal modes alone.

For a conformally coupled scalar in four dimensions, the logarithmic ultraviolet divergence has coefficient $\mathrm{dS}_{d+1}$4 with $\mathrm{dS}_{d+1}$5, consistent with the Weyl anomaly. For pure gravity, the notes quote the semiclassical expansion

$\mathrm{dS}_{d+1}$6

The phase is not discarded: it arises from the finite number of negative conformal modes in the Euclidean gravitational integral. In four dimensions, the trace fluctuation has one negative $\mathrm{dS}_{d+1}$7 mode and five negative $\mathrm{dS}_{d+1}$8 modes, producing a Polchinski-type phase. Different saddle topologies can carry different numbers of negative modes and therefore different phases.

The conformal mode problem is stated plainly. In Euclidean signature the trace part of the metric has a wrong-sign quadratic action, so the path integral is not Gaussian suppressed along the original real contour. Contour rotation can define the integral, but it introduces phases and requires a prescription. The notes do not claim that this problem has been uniquely solved; instead, they compare several contour choices and show how observer degrees of freedom can alter the phase structure.

Static-patch entropy and observer algebras

The Bunch–Davies state restricted to a static patch satisfies the KMS condition with

$\mathrm{dS}_{d+1}$9

This follows geometrically from the periodicity of the invariant distance under imaginary static time shifts and analytically from the Euclidean regularity condition at the horizon. For free QFT, the global state can formally be represented as a thermofield double of antipodal static patches.

The notes carefully distinguish this formal statement from an actual density-matrix construction. In continuum QFT, horizon factorization fails and the static-patch algebra is type III. Including an observer and imposing the gravitational Hamiltonian constraint leads to a crossed-product algebra, which is argued to be type II$-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$0 under appropriate assumptions. In that enlarged algebra, a trace and entropy can be defined, although the entropy remains ambiguous up to an additive constant. The invariant quantity is the variation of generalized entropy,

$-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$1

This construction gives a specific interpretation of why an observer matters: the observer supplies a clock degree of freedom and converts the modular flow of the static-patch algebra into a physical relational evolution. The corresponding Euclidean observer is a particle traversing a great circle of the sphere. Its transverse negative modes and the lapse contour contribute phases that can cancel the conformal-mode phase. The paper presents this as evidence that the observer is not an auxiliary decoration but part of the definition of the observable algebra and path integral.

de Sitter black holes and horizon comparison

The black-hole section reviews Schwarzschild–de Sitter, Reissner–Nordström–de Sitter, and Kerr–de Sitter geometries. Schwarzschild–de Sitter has a black-hole horizon $-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$2 and a cosmological horizon $-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$3, with generically different temperatures. Thus the static region between them is not in global thermal equilibrium. At the Nariai limit,

$-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$4

and the near-horizon geometry becomes $-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$5. The scalar quasinormal frequencies in this limit are

$-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$6

These frequencies match the $-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$7 representation-theoretic structure after the appropriate rescaling.

Charged and rotating solutions exhibit cold, Nariai, ultracold, and lukewarm limits. The ultracold Reissner–Nordström–de Sitter point satisfies

$-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$8

and has a $-(X^0)^2+\sum_{i=1}^{d+1}(X^i)^2=\ell^2$9 near-horizon geometry.

The comparison with AdS black holes is intentionally nontrivial. Both systems exhibit KMS correlation functions and horizon entropy, but their causal responses differ. AdS black holes are associated with scrambling and the standard chaos bound, whereas the notes summarize proposed de Sitter behavior as anti-scrambling, with candidate Lyapunov exponents differing from the AdS value. The paper does not present a definitive universal result for de Sitter OTOCs; it identifies this as an unresolved distinction rather than treating the analogy with black holes as complete.

Low-dimensional realizations

Two-dimensional models are used as controlled settings for the general problems. Timelike Liouville theory arises when a large positive matter central charge is coupled to two-dimensional gravity. Its Weyl mode has a negative-sign kinetic term, directly realizing the conformal mode problem. In the semiclassical limit $\mathbb{R}^{1,d+1}$0, the sphere has a real constant saddle and nonvanishing but parametrically controlled quantum fluctuations. The notes report that the sphere partition function is finite in the relevant timelike Liouville gravity construction and that higher-loop corrections can be computed.

A significant qualification is the mismatch between timelike Liouville gravity and timelike Liouville CFT. The path integral predicts a nonzero sphere partition function dependence whose third cosmological derivative need not vanish, while the analytically continued timelike DOZZ structure constant for the area operator can vanish. The paper proposes that normalization and analytic-continuation prescriptions resolve the discrepancy, but it does not establish a universally accepted equivalence between the two formulations.

Dilaton gravity models, including dS JT gravity, sine-dilaton gravity, and reductions of Schwarzschild–de Sitter, provide finite-dimensional or topological laboratories. They have no local propagating degrees of freedom; their dynamics reside in zero modes, boundaries, and topology. This makes them analytically tractable but also limits their capacity to model the bulk local degrees of freedom present in higher-dimensional de Sitter gravity.

In three dimensions, pure de Sitter gravity is recast as an $\mathbb{R}^{1,d+1}$1 Chern–Simons theory. The wavefunction on a spatial Riemann surface is identified with a Liouville correlator, leading to the complex Liouville string. Its moduli-space amplitudes admit a dual Hermitian two-matrix model with spectral curve

$\mathbb{R}^{1,d+1}$2

The infinitely many branch points lead to an infinite-branch-point topological recursion. The resulting effective eigenvalue count reproduces the semiclassical de Sitter entropy, including logarithmic corrections and the gravitational phase, at the level summarized in the notes: $\mathbb{R}^{1,d+1}$3 This agreement is presented as evidence for a microscopic realization of the perturbative dS$\mathbb{R}^{1,d+1}$4 partition function, while the finite-$\mathbb{R}^{1,d+1}$5 completion remains unspecified by the spectral curve alone.

Four-dimensional higher-spin de Sitter holography

The four-dimensional section addresses the case most directly connected to cosmology. Ordinary Einstein gravity is perturbatively nonrenormalizable, while higher-spin gravity offers an alternative with an infinite gauge symmetry. The proposed dS/CFT relation maps the Hartle–Hawking wavefunction to a three-dimensional Euclidean $\mathbb{R}^{1,d+1}$6 vector-model partition function. The continuation from AdS to dS corresponds schematically to $\mathbb{R}^{1,d+1}$7, exchanging commuting and anticommuting matter in the vector model.

For type-A minimal higher-spin theory, the bulk contains a conformally coupled scalar and integer-spin highest-depth partially massless fields. The boundary wavefunction is

$\mathbb{R}^{1,d+1}$8

with $\mathbb{R}^{1,d+1}$9 denoting boundary sources for bulk higher-spin fields. The notes emphasize that this is a wavefunction, not automatically a normalized state: one must still specify the measure over boundary data.

A bilocal reformulation introduces collective variables

$\mathrm{SO}(1,d+1)$0

where $\mathrm{SO}(1,d+1)$1 are auxiliary commuting fields. At finitely many boundary points, this bilocal has rank at most $\mathrm{SO}(1,d+1)$2, suggesting that the invariant information accessible after gauging higher-spin symmetry scales with $\mathrm{SO}(1,d+1)$3, parametrically matching the de Sitter entropy.

The one-loop higher-spin sphere partition function is expressed as a sum of bulk-minus-edge characters. Its finite part is

$\mathrm{SO}(1,d+1)$4

A proposed gluing formula relates the four-sphere partition function to an integral of the squared $\mathrm{SO}(1,d+1)$5 wavefunction over higher-spin boundary data. The notes report agreement with the one-loop calculation, but the derivation depends on extrapolating AdS/dS relations and on the still incomplete nonperturbative definition of the higher-spin bulk theory.

Limitations and open questions

The paper is a review and its conclusions vary substantially in status. Classical geometry, free-field representation theory, Bunch–Davies correlators, KMS periodicity, and the Nariai quasinormal spectrum are comparatively controlled. By contrast, the interpretation of the de Sitter entropy, the gravitational inner product, the Euclidean contour, and the microscopic Hilbert space remains conjectural.

Several assumptions are especially consequential. The formal factorization into left and right static-patch Hilbert spaces fails for continuum QFT and must be replaced by algebraic constructions. The type II$\mathrm{SO}(1,d+1)$6 crossed-product description depends on including an observer and imposing gravitational constraints; it is not a derivation of a finite-dimensional de Sitter Hilbert space. The entropy is defined only up to an additive constant in that framework. Similarly, the identification of the sphere partition function with a horizon entropy assumes a particular Euclidean saddle and contour prescription, while additional topologies and negative modes can change phases and subleading contributions.

The notes also leave open whether the various lower-dimensional models describe the same physical theory or merely related analytic continuations. In particular, the timelike Liouville CFT/path-integral mismatch is not fully resolved; the precise nonperturbative completion of the complex Liouville matrix model is not determined by its spectral curve; and the proposed higher-spin gluing formula is established only perturbatively in the discussion presented. The relation between static-patch anti-scrambling, observer algebras, and horizon thermodynamics remains unsettled. Finally, no complete four-dimensional, interacting, nonperturbative quantum-gravity definition is provided.

Conclusion

The paper presents de Sitter space as a setting in which familiar notions—energy, asymptotic states, thermal density matrices, entropy, and holographic boundary conditions—must be reformulated rather than directly imported from Minkowski or AdS physics. Its most concrete results are the representation-theoretic classification of de Sitter fields, the derivation of late-time operator algebras and wavefunctionals, the matching of static-patch quasinormal modes to Harish–Chandra characters, and the systematic organization of Euclidean sphere partition functions into bulk, edge, zero-mode, and conformal-mode contributions. The low-dimensional, matrix-model, and higher-spin examples show how pieces of this structure can be realized explicitly, while also making clear that the central questions—especially the microscopic meaning of the de Sitter entropy and the definition of the gravitational inner product—remain open.