How might the history of the Universe end?

The Universe is expanding with acceleration today. Does that mean “forever”? Answering requires knowing not only the present picture, but whether the properties of matter and dark energy will persist into the distant future.

There is no uniquely established ending to cosmic history. If a positive cosmological constant persists, the standard model leads to continued expansion and a future increasingly poor in accessible sources of work. A Big Crunch and a Big Rip require other, additional assumptions. These are conditional scenarios with different standing, rather than three equally likely predictions. PDG: cosmological solutions and the role of dark energy, §22.1.3–22.1.5.

A model extends the history we have observed

Cosmologists compare models with galaxy distances, supernovae, the distribution of matter and the cosmic microwave background. They then calculate what happens next if the model and its laws remain applicable. The farther a prediction extends beyond the observed range, the less secure it becomes.

Dark energy is particularly important: the component used in general-relativistic models to explain late accelerated expansion. A cosmological constant, with unchanging energy density, is the simplest option. If this description changes, the forecast can change too. The shape of space alone cannot determine the ending independently of its contents. PDG: interpretations of acceleration and alternatives, §28.1–28.2.

Heat death means a lack of useful gradients

Imagine a heat engine operating between hot and cold surroundings. The temperature difference allows work to be extracted. Without a useful difference, the mere presence of energy does not provide an endless source of work. The second law distinguishes conserving energy from being able to use it in a cyclic engine. OpenStax: the second law and heat engines, §4.4.

In the cosmic context, heat death means the gradual depletion of accessible free-energy sources that sustain complex processes. Stars consume nuclear fuel, conventional new-star formation weakens, and remnants cool. With continued expansion, distant unbound structures become ever more separated. This is neither a simultaneous moment when “every lamp switches off” nor a claim that energy simply disappears.

Details of the most remote epochs depend on particle physics, black holes and other processes. A review of the future evolution of stars and remnants explicitly stresses that timescales become less secure as extrapolation grows. Adams and Laughlin: §VI and the preceding evolutionary calculations.

Change the assumption and the result changes

With a persistent positive cosmological constant, expansion continues and approaches exponential growth in the distant future. This does not mean that expansion itself inevitably tears apart every atom or bound stellar system. Their own evolution is a separate question.

A Big Rip is possible in simple models where sufficiently unusual “phantom” dark energy retains its properties. Its energy density increases during expansion; its effect eventually grows strong enough to disrupt bound structures. In a model with a constant equation-of-state parameter below −1, the scale diverges in finite cosmic time. A brief episode of such behaviour is insufficient to declare a Big Rip inevitable. Caldwell, Kamionkowski and Weinberg: assumptions and the Big Rip calculation.

A Big Crunch needs a turnaround: the scale stops growing and begins to decrease. This is possible, for example, in models where a future negative field-potential contribution changes the overall dynamics. Even a spatially flat model with particular contents can recollapse. Merely weakening dark energy, however, does not itself guarantee contraction. Kallosh and Linde: examples with negative potentials, §I and §III.

Assumptions determine the scenario

Continued expansion

If: A positive cosmological constant persists, and familiar physics remains applicable.
Then, in the model: The scale grows without turning around; stellar sources gradually fade and accessible gradients are depleted.
Status: Conditional extrapolation of standard ΛCDM, with no established date for heat death.

Big Rip

If: Positive phantom energy with constant w<−1 dominates and maintains this behaviour.
Then, in the model: Its density grows; in the simple model the scale diverges in finite time and bound structures are disrupted.
Status: A hypothetical scenario, not a consequence of present acceleration alone.

Big Crunch

If: Future contents and dynamics allow a turnaround, for example through a negative field potential.
Then, in the model: Expansion reaches a maximum, after which the scale decreases.
Status: Model dependent; a lower dark-energy density alone is insufficient.

A comparison of conditions and consequences, not measured trajectories, timelines, a complete list of models or equal probabilities. Unbound cosmic distances and the internal structure of bound objects are different problems. Dark energy’s behaviour in the distant future is not established.

Branch · a formulaOptional · you can skip this

Why can energy density behave in different ways?

For a homogeneous component that exchanges no energy with other components and has a constant parameter $w=p/\varepsilon$ , the conservation equation gives:

\frac{\varepsilon(a)}{\varepsilon(a_0)} =\left(\frac{a}{a_0}\right)^{-3(1+w)}.

$a$ is the dimensionless scale factor and $a_0$ its reference value; $\varepsilon$ is energy density in J/m³ and $p$ is pressure in Pa, also J/m³. The ratios in the equation are dimensionless. PDG: equations 22.10 and 22.16.

Double the scale, $a/a_0=2$ . For cold matter with $w=0$ , density falls to 1/8 of its initial value; for a cosmological constant with $w=-1$ , it remains unchanged; for a hypothetical component with constant $w=-4/3$ , it doubles. In the third case the exponent becomes positive, so the usual intuition that a greater volume always means lower density no longer works.

These are examples of component behaviour, rather than three complete cosmological models or three calculated ending dates. If $w$ varies or energy is exchanged, the simple power law needs replacing.

Why a distant calendar should not be copied as fact

The future of stellar remnants depends, among other things, on whether protons can decay. This remains a hypothesis under investigation, rather than an observed law. For example, a Super-Kamiokande search published on 13 October 2026 found no significant signal in the decay channels studied. That constrains particular channels; it does not measure the lifetime of every proton. The Super-Kamiokande result.

For black holes, theory predicts Hawking radiation, but the final state and the role of full quantum gravity remain unknown. Statements such as “after this date the last particle will certainly disappear” therefore require far more assumptions than continued expansion alone. What black-hole evaporation actually predicts.

Branch · reading a predictionOptional · you can skip this

What does a precise headline number conceal?

In a 2025 paper by Luu, Qiu and Tye, one best-fit parameter set for a model with an axion field and a cosmological constant gave a total cosmic lifespan of about 33 billion years. This is not 33 billion years from today: roughly 20 billion years remained in that example. Another parameter estimate in the same paper produced a total lifespan near 40 billion years. The original paper: introduction and §5.

The number depends on the chosen field, potential, sign of the constant and parameter estimates. Even a well-fitted example does not measure a future event. Newer DESI results likewise compare specific descriptions of dark energy against specific data; they do not establish a date for the end of the Universe. DESI DR2 IV, 2026: §VI.4–VI.5.

Present tests refine possible futures

We can test whether dark energy varies over the history already accessible to us, how matter structure grows and whether our gravitational description holds. Each test narrows the range of models. It does not remove the need to justify separately how a model behaves in an unobserved future.

Heat death is thus a scenario of gradually depleted accessible gradients; a Big Rip and a Big Crunch require different conditions. Today’s scientific answer is to explain those conditions and the strength of the evidence, rather than choose the most dramatic ending.