Covariant embedding and string-theory origin of all-order $\alpha'$ O(d,d)-invariant bouncing string cosmologies, and their primordial perturbations

The O(d,d)-duality-invariant formulation of tree-level string cosmology resums all $\alpha'$ corrections into a single function $h(z)$ of rescaled canonical momenta; suitable choices of $h(z)$, supplemented by a non-perturbative dilaton potential $V(\phi)$, yield regular non-singular bounces connecting the perturbative string vacuum to matter-dominated FLRW or de Sitter late-time attractors, with automatic isotropization of anisotropic initial data (Conzinu, Fanizza, Gasperini, Pavone, Tedesco, Veneziano, JCAP 2023). The paper itself flags the foundational questions that remain open: whether the reduced Lagrangians/Hamiltonians/Routhians can be obtained as the dimensional reduction, with controlled spatial-gradient corrections, of a generally covariant parent action; whether the phenomenologically promising $h(z)$ and $V(\phi)$ arise from any specific consistent string theory; and how to compute the adiabatic and tensor perturbation spectra through the initial perturbative phase, the high-curvature bounce, and the final phase, including the roles of the Kalb-Ramond axion and of T-duality-compatible stabilization of internal dimensions.

Background

String theory's low-energy limit contains, besides the spacetime metric, a scalar dilaton $\phi$ (whose exponential sets the string coupling $g_s=e^{\phi}$) and an antisymmetric two-form, the Kalb-Ramond field $B_{\mu\nu}$. The parameter $\alpha'$, the squared string length, controls the higher-derivative corrections to the two-derivative effective action. The pre-big bang scenario (Gasperini and Veneziano, 1993) proposes that the universe started in a perturbative string vacuum - nearly flat spacetime and vanishing coupling - and evolved through a phase of growing curvature and dilaton (dilaton-driven inflation) toward a high-curvature regime; the open problem since the 1990s has been the 'graceful exit': crossing the would-be big-bang singularity into a standard decelerated Friedmann-Lemaitre-Robertson-Walker (FLRW) phase.

A key structural fact is that when the background is spatially homogeneous with $d$ commuting translational symmetries (a Bianchi-I metric), the string-frame cosmological equations are invariant under the continuous group $O(d,d)$ of T-duality transformations, which act on the $d$ scale factors and include scale-factor duality $b_i\to 1/b_i$ combined with dilaton shifts; the natural dynamical variable is the shifted dilaton $\bar{\phi}=\phi-\ln(\text{volume})$, which measures the growth of both curvature and coupling. Hohm and Zwiebach (2019) showed that, at tree level in the string-loop expansion but to all orders in $\alpha'$, the $O(d,d)$-invariant cosmological equations can be written in a first-order-in-time form governed by a single function - $F(H_i)$ of the Hubble rates, or after a partial Legendre (Routhian) transformation, $h(z_i)$ of rescaled momenta $z_i$ - into which all higher-derivative corrections are packaged. Smooth analytic $h(z)$ generically gives singular solutions, but special non-analytic choices give regular bouncing solutions; Wang, Wu, Yang and Ying (2019-2020) found the first such solutions by trial and error, and Gasperini and Veneziano (2023) gave a Hamiltonian/Routhian reformulation with a sufficient criterion for regularity, generating a large class of non-singular pre-big bang scenarios.

Conzinu, Fanizza, Gasperini, Pavone, Tedesco and Veneziano (JCAP 12 (2023) 019) then added a non-perturbative, $O(d,d)$-breaking dilaton potential $V(\phi)$, assumed to be sub-Planckian and to have a local minimum at moderate coupling. Imposing 'Asymptotic Past Triviality' (initial data in the far past approach the perturbative string vacuum, where the potential is negligible), they found numerically and analytically that the potential does not spoil the regular bounce, and that the late-time cosmology falls into classes: the dilaton runs back to the vacuum, is reflected (time-reversed evolution), or - most interestingly - is stabilized at a minimum of an effective potential, yielding either a matter-dominated decelerated FLRW universe or a de Sitter-like accelerated phase (de Sitter space being the maximally symmetric solution with positive cosmological constant). Remarkably, whenever the dilaton is stabilized, the attractor is isotropic: the higher-derivative corrections wash out initial anisotropy, removing an old objection to pre-big bang cosmology.

All of this, however, lives in minisuperspace: spatially homogeneous backgrounds with $d$ abelian isometries, no Kalb-Ramond field, and frozen internal compact dimensions. The natural candidate covariant frameworks exist - double field theory (DFT), a reformulation of the NS-NS sector on doubled coordinates with manifest $O(D,D)$ symmetry, and its exactly duality-invariant higher-derivative extension - but the standard DFT is two-derivative, and the exactly invariant extension has not been shown to reproduce the $\alpha'$ expansion of any actual string theory. The 2023 paper therefore ends by identifying the foundational gaps that stand between these phenomenological reduced models and fully embedded, predictive string cosmologies.

Problem Statement

Can the all-order-$\alpha'$, $O(d,d)$-invariant string-cosmology models that realize a regular bounce from the perturbative string vacuum to a matter-dominated FLRW or de Sitter late-time attractor - i.e., the reduced Lagrangians/Hamiltonians/Routhians built from a chosen function $h(z)$ and a non-perturbative dilaton potential $V(\phi)$ - be given a fully embedded string-theoretic foundation? Concretely, the question posed by Conzinu et al. (JCAP 12 (2023) 019) has four linked parts serving this single objective:

1. Covariant parent action. Determine whether a generally covariant action - including the Kalb-Ramond field $B_{\mu\nu}$ - exists whose consistent dimensional reduction to spatially homogeneous backgrounds with $d$ abelian isometries reproduces the reduced dynamics implementing the regular bounce, with controlled higher-derivative ($\alpha'$) and gradient corrections, and construct it (at least perturbatively in spatial gradients) if it does, or establish that it is obstructed.

2. String-theory realisation. Determine whether the specific duality-invariant models that are phenomenologically promising - the viable choices of $h(z)$ together with a $V(\phi)$ capable of dilaton stabilization - arise from any consistent string theory, including loop and non-perturbative effects beyond the phenomenologically postulated potential, or establish that they are obstructed.

3. Perturbations and observables. Within the embedded theory, compute the spectra of adiabatic scalar and tensor cosmological perturbations produced through the three phases - the initial perturbative dilaton-driven-inflation regime (initial conditions can be set in the asymptotically trivial past, avoiding a trans-Planckian problem), the short high-curvature bounce, and the final FLRW or de Sitter phase - and decide whether the standard axion/curvaton mechanism for scale-invariant curvature perturbations (in which a light axion-like field, rather than the background dynamics alone, sources the observed curvature perturbation) is still needed or is modified in this context.

4. Internal dimensions. Implement a stabilization mechanism for the internal compact dimensions compatible with $T$-duality, so that the effective $d$-dimensional description is dynamically grounded, or establish that such a $T$-duality-compatible stabilization is obstructed.

The common scientific objective is to turn the phenomenological reduced bouncing models into fully embedded string-theoretic cosmologies with predictive perturbation theory; part 1 is load-bearing, since the covariant action is the prerequisite for a first-principles computation of perturbations and is expected to constrain the admissible forms of $h(z)$, thereby feeding back into part 2. All four parts are jointly posed toward this objective, part 1 being the prerequisite for parts 2-4. If, however, part 1 or part 2 is established to be obstructed, that obstruction is itself a decisive resolution of the task, and parts 3-4 become moot.

References

1. P. Conzinu, G. Fanizza, M. Gasperini, E. Pavone, L. Tedesco, G. Veneziano, "From the string vacuum to FLRW or de Sitter via alpha' corrections", JCAP 12 (2023) 019. DOI: 10.1088/1475-7516/2023/12/019; arXiv:2308.16076, https://arxiv.org/abs/2308.16076
2. M. Gasperini, G. Veneziano, "Non-singular pre-big bang scenarios from all-order alpha' corrections", JHEP 07 (2023) 144. DOI: 10.1007/JHEP07(2023)144; arXiv:2305.00222, https://arxiv.org/abs/2305.00222
3. O. Hohm, B. Zwiebach, "Duality invariant cosmology to all orders in alpha'", Phys. Rev. D 100 (2019) 126011. DOI: 10.1103/PhysRevD.100.126011; arXiv:1905.06963, https://arxiv.org/abs/1905.06963
4. O. Hohm, B. Zwiebach, "Non-perturbative de Sitter vacua via alpha' corrections", Int. J. Mod. Phys. D 28 (2019) 1943002. DOI: 10.1142/S0218271819430028; arXiv:1905.06583, https://arxiv.org/abs/1905.06583
5. P. Wang, H. Wu, H. Yang, S. Ying, "Non-singular string cosmology via alpha' corrections", JHEP 10 (2019) 263. DOI: 10.1007/JHEP10(2019)263; arXiv:1909.00830, https://arxiv.org/abs/1909.00830
6. P. Wang, H. Wu, H. Yang, S. Ying, "Construct alpha' corrected or loop corrected solutions without curvature singularities", JHEP 01 (2020) 164. DOI: 10.1007/JHEP01(2020)164; arXiv:1910.05808, https://arxiv.org/abs/1910.05808
7. G. Veneziano, "Recent progress in classical string cosmology", Gen. Rel. Grav. 57 (2025) 151. DOI: 10.1007/s10714-025-03482-2, https://link.springer.com/article/10.1007/s10714-025-03482-2
8. T. Codina, O. Hohm, D. Marques, "An alpha'-complete theory of cosmology and its tensionless limit", Phys. Rev. D 107 (2023) 046023. DOI: 10.1103/PhysRevD.107.046023; arXiv:2211.09757, https://arxiv.org/abs/2211.09757
9. O. Hohm, A. F. Pinto, "Cosmological Perturbations in Double Field Theory", JHEP 04 (2023) 073. DOI: 10.1007/JHEP04(2023)073; arXiv:2207.14788, https://arxiv.org/abs/2207.14788
10. S. Angus, S. Mukohyama, "Perturbations in O(D,D) string cosmology from double field theory", Eur. Phys. J. C 85 (2025) 173. DOI: 10.1140/epjc/s10052-025-13841-7; arXiv:2408.13032, https://arxiv.org/abs/2408.13032
11. M. R. Garousi, "O(9, 25) symmetry of heterotic string theory at orders alpha', alpha'^2", JHEP 09 (2024) 150. DOI: 10.1007/JHEP09(2024)150; arXiv:2407.13223, https://arxiv.org/abs/2407.13223
12. A. Pahlavan, M. Ameri, M. R. Garousi, "Heterotic string couplings at order alpha'^3 in NS-NS sector", arXiv:2608.21798 (2026), https://arxiv.org/abs/2608.21798
13. S. Ying, "Large D gravity and low D string via alpha' corrections", JHEP 09 (2024) 156. DOI: 10.1007/JHEP09(2024)156; arXiv:2407.18179, https://arxiv.org/abs/2407.18179
14. Y. Liu, A. Padilla, P. M. Saffin, R. G. C. Smith, "de Sitter vacua in O(d,d) invariant cosmology", Phys. Rev. D 110 (2024) 063522. DOI: 10.1103/PhysRevD.110.063522; arXiv:2404.15401, https://arxiv.org/abs/2404.15401
15. A. S. Arapoğlu, S. Çağan, A. Çatal-Özer, "Stability analysis of the cosmological dynamics of O(D,D)-complete stringy gravity", Eur. Phys. J. C 84 (2024) 848. DOI: 10.1140/epjc/s10052-024-13213-7; arXiv:2405.07825, https://arxiv.org/abs/2405.07825
16. P. Conzinu, G. Fanizza, M. Gasperini, E. Pavone, L. Tedesco, G. Veneziano, "Constraints on the Pre-Big Bang scenario from a cosmological interpretation of the NANOGrav data", JCAP 02 (2025) 039. DOI: 10.1088/1475-7516/2025/02/039; arXiv:2412.01734, https://arxiv.org/abs/2412.01734
17. P. Conzinu, M. Gasperini, E. Pavone, "A simple example of 'non-minimal' Pre-Big Bang scenario", JCAP 11 (2025) 027. DOI: 10.1088/1475-7516/2025/11/027; arXiv:2506.00995, https://arxiv.org/abs/2506.00995
18. M. Gasperini, "Cosmological horizons in regular bouncing backgrounds", arXiv:2605.13636 (2026), https://arxiv.org/abs/2605.13636