How might the universe end?

The universe’s future depends on what fills it and how those ingredients change. In the standard spatially flat model, if dark energy remains a positive cosmological constant, large-scale expansion continues and ordinary matter becomes more dilute. Other assumptions allow other futures. Observing acceleration today does not tell us, by itself, which description will remain true forever.

Expansion is a history, not a countdown

The scale factor compares separations between ideal markers carried with the large-scale cosmic expansion. It does not prescribe the size of every bound galaxy, planetary system or atom. A model of the future must specify both the expansion law and its range of validity. The expansion article develops that distinction.

Cosmological models connect expansion to matter, radiation, spatial curvature and dark energy through general relativity. A positive cosmological constant is one possible description of dark energy: its energy density stays constant as space expands. This is a physical assumption, not the claim that astronomers have identified its microscopic origin. Carroll 2001, §§1.2–1.3 and 2.1

Ask what happens to three ingredients when a freely expanding cosmic region doubles in each direction:

Ingredient	Change in that region	Change in energy density
Nonrelativistic matter, with stable particle masses and conserved particle number	The same particles occupy eight times the volume	Falls to one-eighth
Freely propagating radiation, with conserved photon number	Eight times the volume; each photon also loses half its energy through cosmological redshift	Falls to one-sixteenth
A positive cosmological constant	Its defining density remains unchanged	Stays the same

These are idealized conservation comparisons, not measurements of a future epoch. They explain why a constant component can become increasingly important even without its density increasing.

Three conditional futures

Continued expansion with a positive cosmological constant. If this description persists and matter dilutes away, the scale factor approaches exponential growth. Distant unbound regions separate ever more; existing bound systems do not simply follow that scale factor. Stars still have finite fuel supplies. The idea of eventually exhausting accessible sources of useful work is called heat death. That thermodynamic question requires assumptions about long-term physical processes as well as expansion; it is not a claim that all processes stop at one date. Adams and Laughlin 1997, §VI.D, entropy and heat death

A finite-time runaway in a particular model. A hypothetical component called constant-w phantom energy grows in density as the universe expands. If that behavior persists indefinitely in the classical homogeneous model, expansion reaches a singular limit in finite cosmic time, often called a Big Rip. This conclusion depends on extrapolating that special energy law. It is not a consequence of acceleration alone. Caldwell, Kamionkowski and Weinberg 2003, pp.1–2

A turnaround and contraction. Other histories are possible if the expansion-driving ingredients change. For example, a negative cosmological constant can eventually overcome the diluting positive matter term in an initially expanding model, producing a turnaround and contraction. The sign and evolution of the energy terms matter; the word “gravity” alone does not establish a future collapse. Carroll 2001, §2.2, discussion of recollapse

These examples are not three equally likely forecasts. They expose the assumptions a forecast needs. A classical singularity also marks the failure of that description, not a complete account of unknown physics beyond it.

Try the mathematics: keep the growth law explicit

Preparation: ratios and powers; the final example introduces an exponential. Write $\epsilon$ for energy per volume and $p$ for pressure, which has the same physical units. Define the dimensionless ratio $w=p/\epsilon$. For a separately conserved homogeneous component with constant $w$, energy conservation gives

$$\frac{\epsilon(a)}{\epsilon(1)}=a^{-3(1+w)}.$$

Here $a=1$ at the comparison’s starting point. The factor of three comes from volume growing as the cube of a length. One way to obtain the relation is to insert $p=w\epsilon$ into $d(\epsilon a^3)=-p\,d(a^3)$ and integrate. Carroll 2001, §2.1, fluid-density scaling

At $a=2$, matter with $w=0$ has density $2^{-3}=1/8$ of its initial value. A cosmological constant with $w=-1$ has ratio 1. An invented constant $w=-1.5$ has ratio $2^{1.5}\approx2.83$. The last value illustrates a phantom assumption; it is not a measured value for our future universe.

For constant positive dark-energy density dominating a spatially flat universe, the Hubble parameter approaches a positive constant $H_\Lambda$. In the limiting description, the equation $\dot a/a=H_\Lambda$ gives

$$\frac{a(t)}{a(t_0)}=e^{H_\Lambda(t-t_0)}.$$

After a time $1/H_\Lambda$, the factor is $e\approx2.718$; after twice that interval it is $e^2\approx7.389$. The exponent is dimensionless. No finite time makes this exponential infinite. That is why this idealized constant-density future differs from a Big Rip.

What observations can decide

Measurements test expansion across the past light cone, with an assumed model connecting distances, redshifts and structure. They can challenge a cosmological constant or constrain changing-dark-energy descriptions. They cannot directly observe the remote future.

For example, the DESI collaboration’s 2025 analysis reported that its baryon acoustic oscillation results alone were consistent with the standard cosmological model, while certain combinations with other measurements preferred evolving dark energy. The preference depends on the datasets and model being compared. That is evidence to investigate with further data and systematic checks, not a measurement of a cosmic end date. DESI Collaboration 2025, §VII.1 and Table VI

Test a headline about the end

A headline says: “Expansion accelerates, so galaxies, planets and atoms must all be torn apart.” Which missing assumption changes the conclusion?

Answer and explanation

Acceleration does not by itself require an ever-growing dark-energy density or a finite-time singularity. A persistent positive cosmological constant is a counterexample: it gives accelerating late-time expansion without that runaway. A Big Rip needs additional assumptions about the energy law and its indefinite validity. Bound objects also require their own local dynamics; a cosmic scale factor is not an instruction to enlarge everything.

For classrooms and clubs

Preparation: ratios and the difference between observation and extrapolation. Objective: attach a forecast to its assumptions. In a 10-minute demonstration, double the side of an imaginary box and compare matter, radiation and constant energy density using the table. For a 20-minute investigation, give groups the three future scenarios and ask each to write “If …, then …” followed by an observation that could challenge its starting model. Paper cards provide a complete indoor, no-device version.

Look for the distinction between a model prediction and a measured event. The table’s correct ratios are 1/8, 1/16 and 1; acceleration alone does not select a Big Rip. Do not ask learners to rank scenario probabilities that the supplied evidence cannot determine.

References

• Adams, Fred C., and Gregory Laughlin. “A Dying Universe: The Long-Term Fate and Evolution of Astrophysical Objects.” Reviews of Modern Physics 69, 337–372 (1997). DOI: 10.1103/RevModPhys.69.337. Open manuscript.
• Caldwell, Robert R., Marc Kamionkowski and Nevin N. Weinberg. “Phantom Energy and Cosmic Doomsday.” Physical Review Letters 91, 071301 (2003). DOI: 10.1103/PhysRevLett.91.071301. Open manuscript, PDF.
• Carroll, Sean M. “The Cosmological Constant.” Living Reviews in Relativity 4, 1 (2001). DOI: 10.12942/lrr-2001-1. Open manuscript.
• DESI Collaboration (M. Abdul-Karim et al.). “DESI DR2 Results II: Measurements of Baryon Acoustic Oscillations and Cosmological Constraints.” Physical Review D 112, 083515 (2025). DOI: 10.1103/tr6y-kpc6. Open manuscript, version 3.