🔭 astrophysics

The Case for Space: Estimating Precise Time Delays from Ground- and Space-Based Observations of Lensed Supernovae with Glimpse

This paper introduces the Glimpse model, a novel framework that combines ground- and space-based observations to precisely estimate time delays from unresolved and resolved gravitationally lensed supernovae, demonstrating that such measurements can achieve the accuracy required for competitive $H_0$ constraints in the Rubin-LSST era.

Original authors: Erin E. Hayes, Suhail Dhawan, Stephen Thorp, Justin D. R. Pierel, Nikki Arendse

Published 2026-02-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Erin E. Hayes, Suhail Dhawan, Stephen Thorp, Justin D. R. Pierel, Nikki Arendse

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Cosmic Stopwatch: How We'll Measure the Universe's Speed with "Glimpse"

Imagine the universe is a giant, expanding balloon. For decades, scientists have been arguing about exactly how fast that balloon is inflating. One group of scientists (looking at the baby universe) says it's going one speed, and another group (looking at the nearby universe) says it's going a faster speed. This disagreement is called the "Hubble Tension," and solving it could mean we need to invent new laws of physics.

To settle this fight, we need a super-accurate cosmic stopwatch. Enter Gravitationally Lensed Supernovae (glSNe).

The Setup: The Cosmic Funhouse Mirror

When a massive galaxy sits between us and a distant exploding star (a supernova), its gravity acts like a funhouse mirror. It bends the light from the explosion, creating multiple images of the same event.

Here's the catch: The light takes different paths to get to us. One path is short and direct; another is long and winding. Because of this, the images of the explosion don't appear at the same time. One might flash today, and the other might flash a week later.

The Time Delay: The gap between these flashes is the "time delay." If we can measure this gap perfectly, we can calculate the expansion rate of the universe ( $H_0$ ) with incredible precision.

The Problem: The Blurry Photo

The Vera C. Rubin Observatory (a giant new telescope) is about to discover dozens of these events. But there's a problem: many of these "mirrored" explosions are so close together that the Rubin telescope sees them as one blurry dot, not separate images. It's like trying to time two runners crossing a finish line when they are running so close together you can only see a single blob of motion.

If you can't separate the runners, you can't time them accurately. And if you can't time them, you can't solve the Hubble Tension.

The Solution: Meet "Glimpse"

This paper introduces a new computer model called Glimpse (GausSN Light curve Inference of Magnifications and Phase Shifts, Extended). Think of Glimpse as a super-smart detective that can solve the mystery even when the evidence is blurry.

Here is how Glimpse works, using a simple analogy:

1. The Template (The Script)
We know how a standard supernova explosion should look. It's like a movie script. Glimpse has this script memorized. It knows exactly how the light should rise and fall.

2. The Noise (The Microlensing and Dust)
In the real world, the "movie" gets distorted.

Microlensing: Tiny stars in the foreground galaxy act like extra lenses, making the light flicker randomly.
Dust: Cosmic dust acts like a dirty window, dimming and reddening the light.
Glimpse treats these distortions like a Gaussian Process—a fancy way of saying it uses a flexible rubber sheet to model the wiggles and bumps in the data. It learns to separate the "script" (the real explosion) from the "noise" (the dust and flickering stars).

3. The Magic Trick (Unresolved Data)
This is the paper's big breakthrough. Usually, if you have a blurry dot (unresolved data), you can't measure the time delay. But Glimpse can combine:

The Blurry Dot: Data from the Rubin telescope (which sees the combined light).
The Sharp Photos: A few high-resolution photos from space telescopes (like Hubble or JWST) or deep ground telescopes.

Glimpse takes the blurry dot and the sharp photos and fits them together simultaneously. It's like having a low-resolution video of a dance and a few high-definition snapshots of the dancers' feet. By combining them, Glimpse can reconstruct the exact timing of the dance steps, even if the video is blurry.

The Strategy: How to Get the Best Results

The authors simulated thousands of scenarios to figure out the best way to follow up on these discoveries. They found that the "recipe" for success depends on how bright and clear the image is:

If the image is blurry (Unresolved): You need a lot of help. You need 6 to 8 visits from a space telescope (like Hubble or JWST) looking through different colored filters (like taking photos in red, blue, and infrared). This gives the model enough data to "un-blur" the timing.
If the image is clear (Resolved): You need less help. Just 4 to 6 visits from space or even deep ground telescopes are enough to get a precise time.

The "Sweet Spot":
For the faintest, hardest-to-see supernovae, the paper suggests a "Hybrid Strategy":

Let the Rubin telescope watch the event continuously (high frequency, but blurry).
Send a space telescope (Hubble/JWST) for a few high-quality visits to get sharp details.
Send a giant ground telescope (8-meter class) for deep, high-quality visits to get more data points.

Why This Matters

The paper concludes that with this strategy, we can measure the time delay with an error of less than 0.5 to 0.8 days.

Why does that matter?

If the time delay is 10 days, a 0.5-day error is a 5% precision.
This is the "Goldilocks" zone needed to solve the Hubble Tension.
The authors predict that within just three years of the Rubin telescope starting its work, we could have enough of these precise measurements to finally tell us the true speed of the universe's expansion.

The Bottom Line

We are about to be flooded with data from the Rubin telescope. Most of these cosmic explosions will look like single, blurry dots. But thanks to the Glimpse model, we don't need to wait for perfect, sharp images. By combining blurry ground data with a few sharp space snapshots, we can act as cosmic timekeepers, finally settling the debate on how fast our universe is growing.

It's like solving a puzzle where half the pieces are missing and the other half are blurry, but you have a super-computer that knows exactly what the picture should look like. With Glimpse, we can fill in the gaps and see the whole picture clearly.

1. Problem Statement

Gravitationally lensed supernovae (glSNe) offer a unique, independent probe of the Hubble constant ( $H_0$ ) to resolve the current tension between early-universe (CMB) and late-universe (distance ladder) measurements. However, the precision of $H_0$ estimates derived from glSNe is currently limited by two main factors:

Rarity and Resolution: While the Vera C. Rubin Observatory's Legacy Survey of Space and Time (Rubin-LSST) is expected to discover tens of galaxy-scale glSNe annually, many will be unresolved (multiple images blended into a single point source) due to the compact nature of galaxy-scale lenses and the angular resolution limits of Rubin-LSST.
Modeling Degeneracies: Precise time-delay estimation requires breaking the "mass-sheet degeneracy" (MSD) in lens modeling. This often relies on knowing the absolute magnification of the source, which is difficult to constrain without high-resolution follow-up and accurate modeling of dust extinction and microlensing.

Existing methods for time-delay estimation (e.g., sntd, GausSN, BayeSN-TD) have largely focused on resolved light curves or purely unresolved data, but not the simultaneous modeling of both. Furthermore, they often struggle to account for chromatic microlensing and differential dust extinction across multiple images.

2. Methodology: The Glimpse Model

The authors introduce Glimpse (GausSN Light curve Inference of Magnifications and Phase Shifts, Extended), a Bayesian framework designed to estimate time delays by simultaneously modeling resolved and unresolved observations.

Core Architecture:
- Template Mean Function: Uses SN light curve templates (specifically SALT3 for SNe Ia) to describe the underlying latent light curve $f(t)$ .
- Microlensing Treatment: Models time-varying magnification $\epsilon_M(t)$ using a 2D Gaussian Process (GP). This GP captures smooth variations in time and wavelength (chromatic microlensing), treating microlensing as a perturbation to the template.
- Unresolved Data Integration: Extends the likelihood function to handle unresolved data where the observed flux is the sum of multiple images: $f_{1+2}(t) = [\epsilon_1(t)f_1(t)] + [\beta_{1,2}\epsilon_2(t)f_1(t-\Delta)]$ . The model constructs a combined covariance matrix that accounts for the summation of fluxes and their respective time delays and magnifications.
- Dust Extinction: Incorporates differential dust extinction in the host and lens galaxies as free parameters, assuming a Fitzpatrick (1999) dust law.
Inference: Uses nested sampling (via dynesty) to sample the posterior distribution of hyperparameters, including time delays ( $\Delta$ ), relative magnifications ( $\beta$ ), SALT3 parameters ( $t_0, x_0, x_1, c$ ), and GP hyperparameters for microlensing.

3. Simulation Framework

To test the efficacy of Glimpse, the authors simulated 1,728 unique light curve scenarios based on:

Systems: Four "base objects" (three doubly-imaged, one quadruply-imaged) drawn from a realistic lens catalog (Wojtak et al. 2019), representing the faint end of Rubin-LSST detectability (peak $i$ -band magnitudes of 22–24 mag).
Observing Strategies:
- Rubin-LSST: Simulated with realistic cadences (rolling and non-rolling) and error models. Systems were simulated as both resolved and unresolved.
- Space-Based Follow-up: Simulated observations from HST (optical and NIR filters) and JWST (NIR filters) with varying epochs (2, 4, or 6) and filter combinations.
- Ground-Based Follow-up: Simulated observations from 2m, 4m, and 8m class telescopes with varying depths and cadences (daily to 5-day).
Model Mismatch: To ensure realism, light curves were simulated using the BayeSN SED model but fitted using the SALT3 template, introducing systematic uncertainty from template mismatch.

4. Key Results

A. Time Delay Precision

Rubin-LSST Alone: Unresolved data yields poor constraints (median uncertainty $\sim$ 6.6 days). Resolved data improves this significantly but still struggles to reach sub-5% precision for short delays ( $\Delta < 10$ days) without follow-up.
HST Follow-up:
- Unresolved Systems: 6 epochs of HST follow-up in 4 filters (2 optical, 2 NIR) reduces time-delay uncertainty to 0.7–1.5 days. For systems with $\Delta \approx 15$ days, this achieves sub-5% precision in ~33% of cases.
- Resolved Systems: 6 epochs of HST follow-up achieves 0.5–0.8 day precision. For systems with $\Delta \approx 15$ days, sub-5% precision is achieved in >70% of cases when combined with deep ground-based follow-up.
JWST Follow-up: 2 epochs of JWST (NIR) follow-up provides significant gains for unresolved, faint systems, often outperforming 2 epochs of HST due to wider wavelength coverage and higher sensitivity. However, adding JWST data to HST data can sometimes increase uncertainty due to template mismatch in the NIR.
Ground-Based Follow-up: Deep ground-based follow-up (8m class, 5-day cadence) is highly effective for resolved systems, achieving precision comparable to HST. However, for unresolved systems, ground-based data alone is insufficient; high-resolution space-based imaging is required to break degeneracies.

B. Absolute Magnification and Mass-Sheet Degeneracy (MSD)

The standard candle nature of SNe Ia allows for the estimation of absolute magnification, which breaks the MSD.
Precision: With 4–6 epochs of HST follow-up (optical + NIR), the model achieves absolute magnitude constraints of ~0.1 mag (approx. 10% precision).
Significance: This level of precision is sufficient to constrain the mass-sheet transformation parameter ( $\lambda$ ) to sub-5%, rendering the MSD a sub-dominant source of error in $H_0$ estimation.
Caveat: Accuracy depends heavily on the SN Ia template used. Model mismatch introduces an error floor of ~0.06 mag in absolute magnitude recovery.

C. Impact of Model Misspecification

The study highlights that as data quality (SNR) increases, sensitivity to template mismatch also increases.
For bright systems, fitting high-SNR HST data with an imperfect template (SALT3 vs. BayeSN) can lead to systematic biases in time-delay estimates.
However, the Glimpse model remains robust, with uncertainties generally well-calibrated even in the presence of microlensing and model mismatch.

5. Significance and Conclusions

Feasibility of $H_0$ Constraints: The paper demonstrates that galaxy-scale glSNe discovered by Rubin-LSST can provide competitive $H_0$ measurements. A single system with a 10–20 day time delay, observed with the recommended follow-up strategy, can yield a 7% precision $H_0$ estimate.
Optimal Follow-up Strategies:
- Unresolved Systems: Require 6–8 epochs of space-based follow-up (HST or JWST) in 4+ filters (optical + NIR).
- Resolved Systems: Can achieve high precision with 2 epochs of space-based follow-up if the time delay is short ( $\sim$ 5 days), or 6 epochs for longer delays. Combining deep ground-based follow-up (8m class) with space-based follow-up yields the most consistent results.
Timeline: With the projected discovery rates of Rubin-LSST, the authors argue that a 1.3% precision $H_0$ measurement (sufficient to resolve the Hubble tension at $>5\sigma$ ) could be achieved within three years of the survey's operation, provided the recommended follow-up strategies are implemented.
Future Outlook: The work emphasizes the need for improved SN Ia SED models (particularly in the NIR) and the development of de-blending techniques to extract resolved light curves from unresolved Rubin-LSST images, which would further enhance the utility of these systems.

In summary, Glimpse provides a robust statistical framework to extract precise cosmological parameters from complex, multi-resolution datasets, proving that the Rubin-LSST era will revolutionize time-delay cosmography if paired with strategic space- and ground-based follow-up.