---
title: "Pulsar Timing Arrays and the Hellings–Downs Curve Interactive Lab | Sandbox Physics"
description: "How do you tell a bad clock from gravitational waves using pulsars spread over the sky? By how the timing errors of pulsar pairs line up across the sky."
canonical_url: "https://sandboxphysics.com/labs/pulsar-timing-array-correlations/"
language: "en"
page_type: "lab"
alternate_languages:
  zh-CN: "https://sandboxphysics.com/zh/labs/pulsar-timing-array-correlations/"
last_modified: "2026-11-17"
---

# Pulsar Timing Arrays and the Hellings–Downs Curve

**How do you tell a bad clock from gravitational waves using pulsars spread over the sky?** By how the timing errors of pulsar pairs line up across the sky. A bad clock shifts every pulsar the same way, an error in the Solar System model pushes pulsars on opposite sides in opposite directions, and a gravitational-wave background gives the Hellings–Downs pattern: pairs close together agree, pairs about 80 degrees apart disagree and opposite pairs agree a little again.

Drag pulsars across the sky and mix a clock error, a Solar System ephemeris error and a gravitational-wave background. The pair correlations sort them by their angular pattern.

> Interactive simulation: open this page in a web browser to change the parameters and run the model.

Model assumptions

Simulated array: the 67 sky positions are the real NANOGrav 15-year pulsars, but every timing signal is synthetic. Each pulsar is sampled at 192 evenly spaced epochs over 16.03 years with Gaussian white noise; real arrays have gaps, uneven cadence, intrinsic red noise and a fitted timing model that absorbs part of the signal, so the significance here is higher than in real data. The background is a sum of 32 random plane waves in each of the five lowest frequency bins, with Earth and pulsar terms, and a power-law spectrum of index thirteen thirds as expected from supermassive black-hole binaries. The clock error is one shared series with the same spectrum, and the ephemeris error is a random three-dimensional offset of the Solar System barycentre per frequency, also with that spectrum, so only the angular pattern tells the three apart. Pair correlations use the known noise level to weight the frequency bins, and the three-shape fit is an ordinary least-squares fit that ignores the extra scatter the signal itself causes; the quoted signal-to-noise ratios therefore use the noise-only spread. The colour map on the sky is the delay a pulsar in each direction would see from the shared Earth-side part of the signals; dots also include each pulsar’s own pulsar term.

> Interactive charts computed in the browser from the simulated array.

## Three common errors, three sky patterns

**Background** Millisecond pulsars tick so steadily that astronomers can predict each pulse arrival to within about a microsecond over many years. What remains after subtracting the best timing model is the residual. A pulsar timing array watches dozens of such pulsars and looks for residual wander that several pulsars share.

**Why it matters** A shared wander can come from a clock on Earth, from an error in where we think the Solar System’s centre of mass is, or from gravitational waves passing Earth. All three can have the same spectrum. Only how the wander is shared between pulsars in different directions tells them apart.

- **Focus question**: Why do two pulsars about 80 degrees apart see opposite gravitational-wave delays on average?
- **One-sentence intuition**: A gravitational wave stretches space along one direction and squeezes it along the perpendicular one, so its sky pattern is quadrupolar, not uniform like a clock or two-sided like a shifted origin.

### Core mathematical model

#### How one plane wave delays one pulsar

$$
F^{A}(\hat{\Omega},\hat{p})=\frac{1}{2}\,\frac{\hat{p}^{i}\hat{p}^{j}e^{A}_{ij}(\hat{\Omega})}{1+\hat{\Omega}\cdot\hat{p}}
$$

The response of a pulsar in direction $\hat{p}$ to a wave travelling along $\hat{\Omega}$ with polarisation $A$. The Lab sums 32 such waves in each frequency bin; the colour on the sky is this response summed over all waves.

#### The Hellings–Downs curve

$$
\Gamma(\zeta)=\frac{1}{2}-\frac{x}{4}+\frac{3}{2}\,x\ln x,\qquad x=\frac{1-\cos\zeta}{2}
$$

Averaging the product of two responses over all wave directions gives this curve for two different pulsars separated by $\zeta$ (NANOGrav 15-year paper, eq. 4). It starts at one half, crosses zero near 49 and 122 degrees and bottoms out near 82.5 degrees.

#### Three shapes in one correlation

$$
\langle r_a r_b\rangle\propto A_{\mathrm{gw}}^{2}\,\Gamma(\zeta_{ab})+A_{\mathrm{clk}}^{2}+A_{\mathrm{eph}}^{2}\cos\zeta_{ab}
$$

A clock error is the same in every pulsar, so it adds a constant. A random shift of the Solar System barycentre $\vec{e}$ changes each residual by $\vec{e}\cdot\hat{p}/c$, which correlates pulsars as the cosine of their separation (Tiburzi et al. 2015).

#### Pair estimator and the three-shape fit

$$
\hat{\rho}_{ab}=\frac{\sum_k c_{ak}c_{bk}\,\Phi_k/V_k^{2}}{\sum_k \Phi_k^{2}/V_k^{2}},\qquad \Phi_k\propto f_k^{-13/3}
$$

$c_{ak}$ are the Fourier coefficients of pulsar $a$ at frequency $f_k$, and $V_k$ is the total variance per coefficient. A least-squares fit of the three shapes to all pairs gives the bar chart and the three signal-to-noise readouts.

### Common difficulties

#### A common red process is not yet a gravitational wave

**Typical misconception** If every pulsar shows the same kind of slow wander, the array has detected gravitational waves.

**Better mental model** A clock or ephemeris error can produce the same spectrum. The decisive evidence is the angular pattern between pairs. NANOGrav found a common spectrum with very high confidence but reported the Hellings–Downs pattern as evidence at about three to four sigma, not as a five-sigma detection.

#### Pairs do not sit on the curve one by one

**Typical misconception** With enough data every pair correlation should land on the Hellings–Downs curve.

**Better mental model** The curve is an average over many waves. For one sky of sources, each pair scatters around it even without noise (Allen 2022). Only averages over many pairs follow the curve, which is why the chart bins pairs and why arrays need many pulsars.

### Run the experiment

1. 01

   #### Watch the background pattern

   Leave the default running and watch the colours on the sky, then look at the binned dots in the correlation chart.

   **What to observe:** The sky shows patches of early and late arrival several tens of degrees across. The binned pairs start high, dip below zero around 80 degrees and rise again toward 180 degrees.

2. 02

   #### Swap in a clock or an ephemeris error

   Choose Clock error, then Solar System ephemeris error, and compare the sky and the bar chart with the default.

   **What to observe:** The clock error turns the whole sky one colour at a time; the ephemeris error splits it into two halves. In each case the matching bar grows while the background bar stays near zero.

3. 03

   #### Starve the array

   Choose Only eight pulsars, then drag several pulsars close together on the sky.

   **What to observe:** With only 28 pairs the background signal-to-noise falls to about one, and bunching the pulsars removes the large separations the curve needs.

Physical conclusion

The spectrum says a common signal exists; the angular correlation says what it is.

## Sources and model scope

The Hellings–Downs curve, the power-law spectrum and the pair estimator follow the NANOGrav 15-year background paper, sections 2 and 3. The plane-wave response follows Jenet and Romano (2014), the single-sky scatter of pairs follows Allen (2022) and the ephemeris dipole follows Tiburzi et al. (2015 preprint). The pulsar positions come from the public NANOGrav 15-year data set; an independent Python sky integral checks the curve.

- [NANOGrav Collaboration (2023), The NANOGrav 15-year data set: evidence for a gravitational-wave background, arXiv:2306.16213v1](https://arxiv.org/abs/2306.16213v1)
- [Jenet and Romano (2014 preprint), Understanding the Hellings and Downs curve for pulsar timing arrays in terms of sound and electromagnetic waves, arXiv:1412.1142v1](https://arxiv.org/abs/1412.1142v1)
- [Allen (2022 preprint, v3), Variance in the Hellings–Downs correlation, arXiv:2205.05637v3](https://arxiv.org/abs/2205.05637v3)
- [Tiburzi et al. (2015 preprint), A study of spatial correlations in pulsar timing array data, arXiv:1510.02363v1](https://arxiv.org/abs/1510.02363v1)

[Study the pulsar-timing course: from one ticking star to a Galaxy-sized detector](https://sandboxphysics.com/stories/pulsar-timing-arrays/)

## Related experiments and articles

Pulsar timing arrays and the Hellings–Downs curve · Reading path

### From one ticking pulsar to a Galaxy-sized detector

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

1. [Article **How can a pulsar timing array tell gravitational waves from a bad clock?**](https://sandboxphysics.com/stories/pulsar-timing-arrays/gravitational-waves-or-a-bad-clock/)
2. [Article **What a pulsar timing residual measures**](https://sandboxphysics.com/stories/pulsar-timing-arrays/what-a-timing-residual-measures/)
3. [Article **Deriving the Hellings–Downs curve**](https://sandboxphysics.com/stories/pulsar-timing-arrays/deriving-the-hellings-downs-curve/)
4. [Article **The optimal statistic and its null tests**](https://sandboxphysics.com/stories/pulsar-timing-arrays/optimal-statistic-and-null-tests/)
5. [Article **Recompute a Hellings–Downs statistic from NANOGrav 15-year data**](https://sandboxphysics.com/stories/pulsar-timing-arrays/recompute-the-optimal-statistic/)
6. [Article **Reading the NANOGrav 15-year evidence for a gravitational-wave background**](https://sandboxphysics.com/stories/pulsar-timing-arrays/reading-the-nanograv-15-year-evidence/)

Fields & matter · Starter path

## From a field to an electron microscope

Map an electric field and steer a charged particle. Then find the electron's wavelength and use electron beams to see a crystal and a surface.

[Start here → **Make an Electric Field Visible** Move the probe around a dipole and check that the field always points downhill on the potential landscape. Observe · 8 min](https://sandboxphysics.com/labs/electric-field/)

1. [1 Make an Electric Field Visible](https://sandboxphysics.com/labs/electric-field/)
2. [2 A Charge Bending through Fields](https://sandboxphysics.com/labs/charged-particle/)
3. [3 Electron Wavelength & Voltage](https://sandboxphysics.com/labs/electron-wavelength/)
4. [4 Electron Diffraction & Reciprocal Space](https://sandboxphysics.com/labs/electron-diffraction/)
5. [5 TEM Bright Field, Dark Field & SAED](https://sandboxphysics.com/labs/tem-mode-selector/)
6. [6 SEM Secondary-Electron Topography](https://sandboxphysics.com/labs/sem-topography/)

[View the full path →](https://sandboxphysics.com/topics/fields-and-matter/#starter-path)
