Skip to main content
Sandbox Physics

One Sky and Cosmic Variance

Why does a perfect map still leave uncertainty about the Universe? A sky map samples a finite number of patterns at each angular scale. Even a perfect telescope sees only this one realization, so its measured power fluctuates around the underlying spectrum. Large patterns have fewer independent modes and fluctuate more.

Rotate a synthetic microwave sky and compare its measured angular power with the spectrum that generated it. Draw another sky, or blur the same one with a wider telescope beam.

Interactive modelOne Sky and Cosmic Variance
Filtered sky RMS45.3 μK45.3\,\mu\mathrm K
Selected power / theory0.4170.417
Intrinsic fractional scatter53.5%53.5\%
Independent modes at this degree77

One realization wanders around the theory

Synthetic full-sky record

Angular power

Repeated skies at the selected scale

At each degree, the independent harmonic modes set the scatter. The blue interval contains ninety-five percent of repeated skies under this input spectrum.

Exact sampling relation
(2ℓ+1)C^ℓCℓ∼χ2ℓ+12,σ(C^ℓ)Cℓ=22ℓ+1\frac{(2\ell+1)\widehat C_\ell}{C_\ell}\sim\chi^2_{2\ell+1},\qquad \frac{\sigma(\widehat C_\ell)}{C_\ell}=\sqrt{\frac{2}{2\ell+1}}
0.241<C^3/C3<2.290.241<\widehat C_3/C_3<2.29

The record saves the seed, parameters, complex coefficients for nonnegative orders, filtered coefficients, spectra and interval convention. Negative orders follow the conjugacy relation required by a real temperature field.

Knox (1995), full-sky mode-counting limit

Continue from angular patterns to cosmological inference

One Sky and Cosmic Variance

BackgroundA theory predicts the average strength of temperature patterns across many possible skies. We observe one sky, with one particular set of hot and cold patches.

Why it mattersEven if a telescope measured every patch exactly, its angular-power estimate would still fluctuate because there are only finitely many independent patterns at each scale.

One-sentence intuition
Large angular patterns have fewer independent modes, so one sky gives a less precise estimate of their average power.

Core mathematical model

Temperature patterns on the sphere

ΔT(n^)=∑ℓ=2ℓmax⁡∑m=−ℓℓaℓmYℓm(n^),aℓ,−m=(−1)maℓm∗\Delta T(\hat n)=\sum_{\ell=2}^{\ell_{\max}}\sum_{m=-\ell}^{\ell}a_{\ell m}Y_{\ell m}(\hat n),\quad a_{\ell,-m}=(-1)^m a_{\ell m}^*

Orthonormal complex spherical harmonics describe angular patterns. The conjugacy relation makes the temperature real. The monopole and dipole are excluded, and all temperatures are in microkelvin.

Estimating power from one sky

C^ℓ=12ℓ+1∑m=−ℓℓ∣aℓm∣2\widehat C_\ell=\frac{1}{2\ell+1}\sum_{m=-\ell}^{\ell}|a_{\ell m}|^2

At each degree, sum the squared coefficient amplitudes and average over the independent real modes. The map and this estimate use exactly the same saved coefficients.

The full-sky Gaussian sampling law

(2ℓ+1)C^ℓ/Cℓ∼χ2ℓ+12,σ(C^ℓ)/Cℓ=2/(2ℓ+1)(2\ell+1)\widehat C_\ell/C_\ell\sim\chi^2_{2\ell+1},\qquad \sigma(\widehat C_\ell)/C_\ell=\sqrt{2/(2\ell+1)}

The squared independent Gaussian coordinates add to a chi-square variable. Its central sampling interval is asymmetric, especially at low angular degree. This is scatter among possible skies at fixed theory.

Common difficulties

Uncertainty survives a perfect measurement

Typical misconceptionA noiseless map must reveal the theoretical spectrum exactly.

Better mental modelA noiseless map gives the exact coefficients of our realization. The theoretical spectrum describes an ensemble of realizations, and the finite number of modes leaves sampling scatter.

Run the experiment

  1. 01

    Keep the theory fixed

    Before pressing Draw another sky, predict whether the white theoretical line or the gold points will move. Draw several skies.

    What to observe: The theory stays fixed and the gold points wander. A low or high point can arise from the finite random sample alone.
  2. 02

    Compare large and small patterns

    Choose A few large patterns, then Small patterns appear. Under Select a scale and a seed, move Selected angular degree from the lowest to a higher degree.

    What to observe: The repeated-sky distribution narrows at higher degree. More independent modes give a more stable estimate of the input power.
  3. 03

    Blur the same sky

    Predict which patches disappear first, then widen Beam width without drawing another sky. Save the sky and harmonics.

    What to observe: Small patches fade fastest. The cyan power curve drops at high degree, and the original gold points stay fixed. The saved coefficients specify both views.

Related experiments and articles

What this page links to directly

See every article series on the content map →

Continue exploring

Testing a model of the universe

How do we test a model of the universe against data, and how far can we trust the result?

  1. LabSupernova Distances and the Distance LadderChange the matter fraction and see which expansion history fits the supernovae.
  2. LabDark Energy or Modified GravityChange gravity’s running strength and compare clustering growth with Einstein gravity.
  3. LabFlat Torus Universe and Matched CirclesChange the box edge and watch the matched circles shrink and vanish as the box grows.
  4. You are hereOne Sky and Cosmic VariancePress Draw another sky a few times and watch the measured power wander around the theory.
  5. LabCosmology, Likelihoods and PriorsWiden the prior half-width and watch whether the data or the prior sets the answer.
  6. LabNeural Posterior Calibration and Model MismatchPick Noise louder than assumed and watch whether the 90% regions still catch the truth.

Cosmology from expansion to inference · Reading path

Expansion, sound, sky statistics and inference

Articles on this path that use this experiment, followed by the other experiments they suggest.

  1. ArticleHow much can one sky tell us about its power?
Something broken, confusing, or missing on this page?Tell us
What kind of feedback?

Sent with this page’s address, its Lab ID, the site language and your device type. Please leave out personal details. Feedback is kept for up to 12 months.Privacy policy