Future Expansion, Recollapse, and Singular Fates

Spatial curvature alone does not determine the universe’s fate. Expansion forever, recollapse and a finite-time singularity depend on the full stress–energy history, including whether an effective equation of state persists, whether vacuum energy is positive or negative, and whether the low-energy description remains valid. Two future integrals give a clean first classification, but model transitions can change it.

Helpful background. Einstein Equations determine the Friedmann constraint, Relativistic Matter and Compact Objects supplies energy-condition language, Causality and Global Structure distinguishes conformal from proper-time endpoints, Relativistic Cosmology supplies component evolution, and Mathematical Relativity clarifies singular and extendible boundaries.

Proper-time and conformal-time classification

Consider a spatially flat universe that becomes dominated forever by one perfect fluid with constant $w=p/\rho$. Conservation and the Friedmann equation give

$$\rho(a)=\rho_\star a^{-3(1+w)}, \qquad H(a)=H_\star a^{-3(1+w)/2},$$

where the positive constants absorb the normalization of $a$. The remaining future proper time and conformal time scale as

$$\Delta t\propto\int_a^\infty (a')^{(1+3w)/2}\,da', \qquad \Delta\eta\propto\int_a^\infty (a')^{(-1+3w)/2}\,da'.$$

The first converges only for $w\lt-1$; the second converges only for $w\lt-1/3$. The comparison axes are remaining proper time, remaining conformal time and the regularity of the endpoint:

Dominant constant $w$	Future proper time	Future conformal time	Idealized fate
$w\gt-1/3$	Infinite	Infinite	Decelerating expansion, no event horizon
$w=-1/3$	Infinite	Logarithmically infinite	Coasting boundary case
$-1\lt w\lt-1/3$	Infinite	Finite	Accelerated power-law expansion with an event horizon
$w=-1$	Infinite	Finite	Asymptotic de Sitter expansion
$w\lt-1$	Finite	Finite	Big-rip endpoint if the constant-$w$ model persists

For $w\lt-1$, density grows as the universe expands and $a\to\infty$ at finite cosmic time. Caldwell, Kamionkowski and Weinberg derive the finite remaining time in the dark-energy-dominated approximation and discuss the destruction of bound systems Caldwell et al. 2003, pp. 2–3 and Table I (PDF). This is a conditional extrapolation of a phantom-fluid model, not a prediction that such a fluid exists or remains valid to arbitrarily high density.

The result can be put in local units without claiming that local data fix the future. If the stated model holds from $t$ onward, direct integration gives

$$t_f-t=\frac{2}{3|1+w|H(t)} \quad (w\lt-1), \qquad D_{\rm e}(t)=\frac{2}{|1+3w|H(t)} \quad (w\lt-1/3).$$

For example, $w=-4/3$ gives a remaining time $2/H$ and an event-horizon proper radius $2/(3H)$. The latter is already smaller than the Hubble radius. The factors depend on the assumed equation of state over the whole future, not just its fitted value at the observation time.

Turnaround and recollapse

With spatial curvature and a cosmological constant retained,

$$H^2=\frac{8\pi G}{3}\rho+\frac{\Lambda}{3}-\frac{K}{a^2}.$$

A turnaround can occur only where the right-hand side reaches zero along the actual solution. Positive curvature $K\gt0$ can enable this for ordinary matter when $\Lambda=0$, but it does not guarantee recollapse: a sufficiently important positive $\Lambda$ keeps the right-hand side positive at large $a$. Conversely, negative vacuum energy can force a turnaround even when $K=0$. At a candidate root one must also check

$$\dot H=-4\pi G(\rho+p)+\frac{K}{a^2};$$

$\dot H\lt0$ identifies a crossing from expansion to contraction rather than a bounce.

A flat dust universe with $\Lambda=-3h^2\lt0$ makes the second possibility explicit. Write $H^2=C/a^3-h^2$, with $C\gt0$ and $h\gt0$. The solution

$$a(t)=a_{\max}\sin^{2/3}\!\left(\frac{3ht}{2}\right), \qquad a_{\max}=\left(\frac{C}{h^2}\right)^{1/3}, \qquad 0\lt t\lt\frac{2\pi}{3h}$$

starts with a big bang, turns around at $t=\pi/(3h)$, and recollapses. At the turnaround $\dot H=-3h^2/2\lt0$. Spatial flatness has not prevented recollapse; the negative cosmological constant supplied the competing term.

The familiar rule “closed means recollapsing” is therefore a special result for restricted matter content, not a geometric theorem. A scalar potential can change sign, a component can decay, or modified high-energy dynamics can invalidate the assumed Friedmann equation. This page compares general-relativistic background fates; theory-specific beyond-GR evolution belongs elsewhere.

Singular endpoints are not all big rips

A big rip has finite remaining proper time with $a$, $H$ and the phantom density diverging. A sudden singularity can instead occur with finite nonzero $a$, finite $H$ and finite $\rho$, while $p$ and $\ddot a$ diverge. Barrow constructs an explicit family

$$a(t)=\left(\frac{t}{t_s}\right)^q(a_s-1) +1-\left(1-\frac{t}{t_s}\right)^n$$

for $0\lt t\lt t_s$, where $a_s\gt1$, $1\lt n\lt2$, and $0\lt q\le1$. As $t\to t_s$, the scale factor and first derivative remain finite but $\ddot a\to-\infty$ Barrow 2004, Eqs. (2)–(14), pp. 1–3 (PDF). The example can retain $\rho\gt0$ and $\rho+3p\gt0$; its failure mode is an unbounded pressure not controlled by a regular equation of state.

This curvature pathology still does not prove that every causal geodesic must end there. FLRW geodesic equations involve $a$ and $\dot a$, which remain finite at $t_s$, whereas curvature involves $\ddot a$. A suitable lower-regularity continuation can therefore continue geodesics even though a $C^2$ extension through the divergent curvature is impossible. The regularity of the proposed continuation must be stated explicitly.

These cases show why three labels must remain separate: endless expansion describes $a(t)$, an event horizon describes a conformal-time integral, and a singular fate concerns extendibility or geodesic length. An asymptotically de Sitter model expands forever yet has an event horizon; a big rip reaches $a=\infty$ in finite proper time; a sudden singularity reaches only a finite $a_s$.

Vacuum transitions and evidential boundary

The analytic classification and model-dependent effective-equation-of-state or vacuum-decay claims here are reviewed through 2026-08-19. A metastable vacuum could decay through a bubble transition and replace the assumed late-time component, but the decay rate, daughter vacuum and global continuation depend on a specified quantum theory and state. No generic Friedmann argument fixes them. Likewise, a fitted effective $w$ over a finite redshift interval cannot be extrapolated automatically to the infinite future.

A later calculation that changes the stability or lifetime of an invoked vacuum, establishes a controlled crossing of $w=-1$, or supplies a new singularity/extension theorem could change this conclusion. Present observations constrain past-light-cone expansion under a model; they do not observe the future endpoint. The robust result is conditional: once $H(a)$ is specified over the whole future, the two integrals and turnaround test determine the corresponding fate.

Exercises

For a flat, single-component model with constant $w=-4/3$, determine whether the future proper-time and conformal-time integrals converge. What conclusion is allowed, and what assumption limits it?

Solution

The proper-time integrand has exponent $(1+3w)/2=-3/2$, and the conformal-time integrand has exponent $(-1+3w)/2=-5/2$. Both integrals to $a=\infty$ converge, so the idealized solution reaches a big-rip endpoint and has a finite event-horizon radius. The conclusion requires the constant-$w$ general-relativistic fluid description to remain valid throughout the future; decay, changing $w$ or new high-energy dynamics can alter it.

References

• Barrow, John D. “Sudden Future Singularities.” Classical and Quantum Gravity 21 (2004): L79–L82. DOI. Open PDF.
• Caldwell, Robert R., Marc Kamionkowski, and Nevin N. Weinberg. “Phantom Energy and Cosmic Doomsday.” Physical Review Letters 91 (2003): 071301. DOI. Open PDF.