Data-driven core collapse supernova multilateration with first neutrino events

This paper presents a data-driven multilateration method that utilizes the arrival times of the first inverse beta decay neutrino events across multiple detectors, combined with a reference lightcurve from a larger detector, to rapidly estimate the direction and generate probability skymaps for Galactic core-collapse supernovae without relying on simulations.

The Gist

Imagine a massive star in our galaxy suddenly collapsing. It's like a cosmic firework that goes off in total silence, but it sends out a massive burst of invisible "ghost particles" called neutrinos minutes before the actual light (the explosion we see with telescopes) ever reaches us.

Scientists want to know exactly where in the sky this explosion happened so they can point their telescopes there immediately. This paper describes a clever, data-driven way to find that location using a network of neutrino detectors around the world, acting like a cosmic GPS.

Here is the breakdown of their method, explained with everyday analogies:

1. The Problem: The "Big Ear" vs. The "Small Ear"

Imagine you and a friend are trying to figure out where a thunderstorm is by listening to the first crack of thunder.

• You have a giant, super-sensitive microphone (a Large Detector like Super-Kamiokande).
• Your friend has a tiny, cheap radio (a Small Detector like LVD or SNO+).

If the storm is far away, your giant microphone will hear the very first tiny rumble almost instantly. Your friend's small radio, however, might miss that tiny rumble and only register the sound once it gets loud enough to trigger the volume knob.

The Bias: Because your friend's radio is "slower" to react to the faint start of the storm, it will always seem like the thunder arrived later at their house than at yours. If you just subtract the two times, you get a wrong answer. You might think the storm is coming from the wrong direction because you didn't account for the fact that your friend's radio is just less sensitive.

2. The Solution: The "Data-Driven Correction"

The authors realized that previous methods relied on complex computer simulations to guess how much "late" the small radio would be. But simulations can be wrong if the storm behaves differently than expected.

Instead, they invented a self-correcting math trick that uses the data itself:

• They take the list of every sound your giant microphone hears.
• They mathematically "shrink" that list to pretend it belongs to the small radio. (Imagine turning down the volume on your recording until it looks like the small radio's recording).
• By comparing the real first sound on the big radio with the simulated first sound on the "shrunken" big radio, they can calculate exactly how much "lag" is caused just by the size difference.

The Metaphor: It's like realizing your friend is late not because the storm is far away, but because they are walking slower. Once you calculate their walking speed based on their own steps, you can subtract that delay and find the true time the thunder started.

3. The Result: A "Sky Map" of Possibilities

Once they fix the timing error, they use the tiny differences in arrival times between detectors (like JUNO in China, Super-K in Japan, LVD in Italy, and SNO+ in Canada) to triangulate the source.

• The Output: They don't get a single pinpoint dot. Instead, they generate a probability map of the sky.
• The Accuracy: It's not perfect. The "68% confidence area" (the zone where the explosion is most likely to be) covers a few thousand square degrees.
  • Analogy: If the whole sky were a giant pizza, this method tells you the explosion is definitely on one slice, but maybe not the exact pepperoni on that slice.
  • Why is this good? Even though it's a big slice, it's a massive improvement over guessing. It gives astronomers a specific region to scan immediately, saving precious minutes before the light of the explosion arrives.

4. Why This Matters

This method is fast and robust.

• Fast: It only needs the time of the first few neutrinos detected. It doesn't need to wait for the whole explosion to finish.
• Robust: It doesn't rely on guessing how the star exploded (the physics models). It just uses the raw data from the detectors.

The Bottom Line:
When a star dies in our galaxy, this new method acts like a rapid-response team. It uses the "first whispers" of neutrinos, corrects for the fact that some detectors are "deaf" to the faintest sounds, and points the world's telescopes in the right direction within minutes. This ensures that when the light finally arrives, we aren't looking at the wrong part of the sky.

Technical Summary

Here is a detailed technical summary of the paper "Data-driven core collapse supernova multilateration with first neutrino events."

1. Problem Statement

A Galactic Core-Collapse Supernova (CCSN) emits a burst of neutrinos minutes to hours before electromagnetic radiation arrives. The SNEWS2.0 network aims to alert the global astronomical community and determine the supernova's location (pointing) using the relative arrival times of these neutrinos at different detectors.

The primary challenge addressed in this paper is the bias inherent in "first-event" multilateration when comparing detectors of vastly different sizes (event yields).

• The Bias: In a larger detector, the first neutrino event is statistically likely to occur earlier in the burst than in a smaller detector, simply due to the higher event rate. If one simply subtracts the time of the first event in Detector A from Detector B ($t_{1,A} - t_{1,B}$), the result is biased, leading to incorrect time lags and, consequently, inaccurate sky localization.
• Limitations of Existing Methods: Previous approaches often relied on Monte Carlo simulations with specific supernova models to estimate these biases and uncertainties. This introduces model dependence and computational overhead, which is undesirable for rapid, real-time alerts.

2. Methodology

The authors propose a data-driven, model-independent method to correct for this bias and estimate uncertainties using only the observed event times.

A. Bias Mitigation (The "Corrected Lag")

The method derives an expectation value for the first event time based on the observed event rate $R(t)$.

1. Expectation Value Calculation: For a detector with events $t_1, t_2, \dots, t_N$, the expectation value of the first event time, $\langle t_1 \rangle$, is approximated as a weighted average that emphasizes early events:
   $$\langle t_1 \rangle \approx \frac{\sum_{j=1}^N t_j e^{-j}}{\sum_{j=1}^N e^{-j}}$$
   Here, the weight $e^{-j}$ approximates the cumulative Poisson probability of having zero events up to time $t_j$.
2. Scaling for Yield Differences: To compare a large detector (A) with a smaller detector (B), the rate of the large detector is scaled by a factor $\alpha < 1$ (where $\alpha$ represents the ratio of expected yields). The expectation value for the smaller detector is estimated using the large detector's data but with the scaling factor:
   $$\langle t_{\alpha} \rangle \approx \frac{\sum_{j=1}^N t_j e^{-\alpha j}}{\sum_{j=1}^N e^{-\alpha j}}$$
3. Corrected Lag ($Z_{AB}$): The bias-corrected time difference is calculated as:
   $$Z_{AB} = t_{1,A} - t_{1,B} - \langle t_{1,A} \rangle + \langle t_{\alpha A} \rangle$$
   This removes the systematic shift caused by the difference in detector size, leaving an estimate of the true geometric time lag.

B. Uncertainty Estimation

The paper derives a formula to estimate the covariance matrix of the time differences directly from the data, avoiding the need for external simulations.

• The variance of the first event time is estimated using the second moment of the weighted distribution: $\langle t^2 \rangle - \langle t \rangle^2$.
• The total uncertainty for a pair of detectors is the quadrature sum of their individual uncertainties.
• These uncertainties are used to construct a $\chi^2$ function to minimize over the sky to find the most likely direction.

C. Binned Data Extension

The authors also provide a formulation (Eq. 9) for detectors that report binned data with significant backgrounds (e.g., IceCube, KM3NeT), allowing these large telescopes to be incorporated into the multilateration network.

3. Key Contributions

1. Analytical Bias Correction: A rigorous derivation of a data-driven correction for the "first-event bias" that depends only on the observed event times and relative yields, eliminating the need for supernova models.
2. Direct Uncertainty Estimation: A method to calculate the covariance matrix and confidence intervals directly from the event data, enabling rapid probability skymaps.
3. Robustness: The method works effectively even when detectors have yields differing by an order of magnitude (e.g., Super-Kamiokande vs. SNO+).
4. Extension to Binned Data: A framework to include high-background detectors that cannot resolve individual events, expanding the network's reach.

4. Results

The method was tested using Monte Carlo simulations with Bollig 2016 supernova models (27 $M_\odot$ and 11.2 $M_\odot$ progenitors) and four operational detectors: Super-Kamiokande (SK), JUNO, LVD, and SNO+.

• Bias Reduction:
  • Uncorrected: For a large/small detector pair (e.g., SK vs. SNO+), the uncorrected bias could exceed 10–15 ms at 10 kpc, growing to 30–40 ms at 20 kpc.
  • Corrected: After applying the data-driven correction, the residual bias was reduced to < 1 ms (specifically < 0.2 ms for similar yields, < 1 ms for disparate yields), representing an improvement of over an order of magnitude.
• Uncertainty Calibration:
  • The estimated uncertainties were found to be slightly lower than the observed RMS. The authors applied a conservative inflation factor of 1.2 to the estimated uncertainties to ensure reliable confidence intervals.
• Sky Localization (Pointing):
  • Confidence Intervals: The 68% confidence interval area varies by direction and distance but is typically 3,000 to 5,000 square degrees for a supernova at 10 kpc.
  • Coverage: The method provides "over-coverage" (conservative), meaning the true supernova location falls within the calculated 68% confidence region approximately 73% of the time (compared to the ideal 68%). Without the bias correction, the coverage drops significantly (to ~38%), often pointing observers in the wrong direction.
  • Geographical Dependence: The skymap structure reflects the northern hemisphere latitude of the detectors (22°–46°), showing larger error ellipses perpendicular to the detector baselines.

5. Significance

This paper provides a critical tool for the SNEWS2.0 network and the multi-messenger astronomy community:

• Speed and Robustness: It enables a rapid, real-time pointing estimate immediately upon the detection of the first few neutrino events, without waiting for full lightcurve analysis or relying on specific supernova models.
• Complementary Role: While elastic scattering events in large detectors (like SK or DUNE) can eventually provide higher precision (3–7 degrees), they require more events and time. This first-event method provides a quick, initial alert to guide telescopes to the correct region of the sky while more precise methods are being computed.
• Inclusivity: By correcting for yield differences and handling binned data, it allows the combination of diverse detector types (water Cherenkov, liquid scintillator, and large-volume telescopes) into a single, coherent localization network.

In summary, the authors have developed a practical, model-independent algorithm that transforms the simple "first-event" timing method from a biased, unreliable tool into a robust, data-driven technique capable of providing accurate initial sky maps for Galactic supernovae.