Dark Energy
Quantum field theory predicts the energy density of empty space. The prediction is wrong by a factor of 10^120 — the largest miss between theory and measurement in the history of physics. The residual, whatever it actually is, makes up 68% of the universe and is pushing galaxies apart faster every year. We call it dark energy because we have no better name.
The 1998 surprise
Before 1998, the open question was how fast gravity was slowing cosmic expansion. Two teams — Saul Perlmutter's Supernova Cosmology Project and the High-Z team of Brian Schmidt and Adam Riess — measured Type Ia supernovae as standard candles out to redshift z ≈ 0.5. Distant supernovae came back dimmer than a decelerating universe permits. Expansion was accelerating. Nobel Prize, 2011.
The fix, mathematically, was already on Einstein's desk. He had inserted Λ — the cosmological constant — in 1917 to keep the universe static, then dropped it in 1929 after Hubble. The 1998 data put it back, but with a value 120 orders of magnitude smaller than what quantum field theory says the vacuum should weigh. ΛCDM (Lambda Cold Dark Matter) became the standard model of cosmology by assuming Λ is exactly that small, exactly constant, and we stop asking why.
The three models on the table
Cosmological constant (Λ). A fixed property of spacetime. Equation of state w = −1 exactly. Universe accelerates forever; ends in the Big Freeze — maximum entropy, no bound structures, everything thermal and isolated.
Quintessence. A scalar field slowly rolling down a potential. w > −1, evolving with time. Parameterized as w(a) = w₀ + wₐ(1−a), where a is the scale factor. Λ corresponds to w₀ = −1, wₐ = 0; anything else is new physics.
Phantom energy. w < −1. Dark energy density grows as space expands, in feedback with itself. Galaxy clusters unbind, then stars, then atoms — the Big Rip. Robert Caldwell proposed it in 2002 as a deliberate provocation. DESI is now making it less of a joke.
What DESI is seeing
The Dark Energy Spectroscopic Instrument at Kitt Peak is building the largest 3D map of the universe ever attempted: 40+ million galaxy and quasar redshifts across 11 billion years, using baryon acoustic oscillations — a 150-megaparsec ripple frozen into the matter distribution from the early universe — as a standard ruler.
- Year 1 (April 2024): ~2.5σ hint that dark energy is not constant.
- DR2 (19 March 2025): 3.1σ from BAO alone. 4.2σ combined with Pantheon+ and DESY5 supernova samples. Best-fit w₀ ≈ −0.77, wₐ ≈ −0.86.
- Dark Energy Survey, independently (2025): 3.2σ in the same direction.
The signal says dark energy was stronger in the past and is weakening toward zero now. It also passes through the phantom divide w = −1 — "phantom crossing" — which most simple scalar field models forbid. Four independent datasets aligning is unusual; the discovery threshold of 5σ has not been crossed. DESI's full five-year dataset is expected in 2027; Euclid's DR1 lands October 2026.
Why the number is the scandal
The vacuum should contribute zero-point energy from every quantum field. Summing the modes up to the Planck cutoff gives roughly 10^113 J/m³. Observation gives roughly 10^−9 J/m³. Pick any cutoff you like — Planck, GUT, electroweak — and the answer is still off by 50 to 120 orders of magnitude. There are three live responses:
- Anthropic selection. In a string-theory landscape of ~10^500 vacua, observers only exist where Λ is small. We measure a small Λ because galaxies are a precondition of measurement. Steven Weinberg predicted the rough scale this way in 1987, before the discovery.
- Hidden cancellation. Some unfound symmetry zeros out the contributions to one part in 10^120. No candidate symmetry has worked.
- Dynamical Λ. It is not a constant at all — it is a field still relaxing. This is what DESI may be showing.
The de Sitter Swampland Conjecture (Obied, Ooguri, Spodyneiko, Vafa, 2018) goes further: stable de Sitter vacua may be forbidden in string theory. If so, a true cosmological constant is impossible and quintessence is mandatory. DESI's hint is exactly the prediction.
What's contested
Whether DESI is seeing physics or systematics. The supernova samples driving the combined 4.2σ — DESY5 and Pantheon+ — have known calibration tensions between low-redshift and high-redshift subsamples; some of the apparent w₀-wₐ signal may be absorbed into those offsets. The w₀wₐ parameterization itself is a Taylor expansion and may simply be the wrong shape to describe whatever dark energy actually does. And the concept hubble tension — the 5σ disagreement between local (SH0ES, ~73 km/s/Mpc) and CMB-inferred (Planck, ~67.4 km/s/Mpc) values of the Hubble constant — could itself be sourced by evolving dark energy, or by something else entirely. Two anomalies arguing about whether they are the same anomaly.
The unknown that nobody has cracked: what is the field, mechanically? Quintessence requires a scalar with mass roughly 10^−33 eV — twelve orders of magnitude lighter than the lightest neutrino bound. No particle physics framework predicts such a field; protecting it from radiative corrections that would push it heavier is its own fine-tuning problem.
Why this has to do with other realms
If dark energy weakens enough to reverse, the universe recollapses. A 2026 calculation by Hoang Nhan Luu and Paul Steinhardt put a Big Crunch at roughly 20 billion years from now if DESI's central values hold. That bounds the timescale available for any concept fermi paradox resolution that depends on civilizations spreading on cosmological scales — the "deep future" suddenly has an end date instead of an asymptote. It also constrains concept arrow of time: a recollapsing universe needs a low-entropy future boundary condition as fine-tuned as the Past Hypothesis, which most cosmologists consider absurd. The same data point reshapes thermodynamics, the fate of intelligence, and the metaphysics of time at once.
An open question
If w₀ ≈ −0.77 holds at 5σ in DESI DR5, the cosmological constant is dead — but quintessence requires a scalar field nobody has any independent reason to expect. Which is the more uncomfortable conclusion: that the universe runs on a particle we cannot detect, or that string theory's landscape is the right framework for picking which one we got?
Key sources
- Perlmutter et al. 1999, ApJ 517 — original Supernova Cosmology Project acceleration paper.
- Riess et al. 1998, AJ 116 — High-Z team companion paper.
- DESI Collaboration 2025, "DESI 2024 VI: Cosmological constraints from BAO measurements" (arXiv:2404.03002) and DR2 follow-up (arXiv:2503.14738).
- Weinberg 1989, Rev. Mod. Phys. 61 — the canonical statement of the cosmological constant problem.
- Obied, Ooguri, Spodyneiko, Vafa 2018, "De Sitter Space and the Swampland" (arXiv:1806.08362).
- Caldwell 2002, Phys. Lett. B 545 — phantom energy proposal.
- To verify: Luu & Steinhardt 2026 Big Crunch timescale paper covered in ScienceDaily.
Further reading
- The 4% Universe by Richard Panek (2011) — the discovery-era history, with the personalities. Worth reading for how contingent the 1998 result felt at the time.
- Sean Carroll's Mindscape episodes on dark energy and the Swampland — accessible without dumbing down.
- concept cosmological natural selection — Lee Smolin's alternative to anthropic reasoning for fine-tuning.
- arXiv:1807.06209 (Planck 2018 cosmological parameters) — the ΛCDM baseline DESI is challenging.
- The First Three Minutes by Steven Weinberg — old but unmatched on how to think about cosmological evidence.
See Also
- concept dark matter — the other 27% no one has found, and the one that might be entangled with dark energy in unified dark-sector models.
- concept hubble tension — the 5σ anomaly that may or may not share a cause with DESI's signal.
- concept holographic principle — dS/CFT and what a positive Λ even means in a holographic universe.
- concept arrow of time — what a recollapsing cosmos demands of the future boundary.
- concept fermi paradox — bounded futures change the long-civilization calculus.
- concept simulation hypothesis — the 10^120 fine-tuning, taken seriously by people who shouldn't.