Steeling Weak Lensing Source Galaxy Samples against Systematics using Wide Field Spectroscopy

This paper demonstrates that weak lensing analyses can maintain their cosmological constraining power using significantly sparser source galaxy samples ($5\,\text{arcmin}^{-2}$) by leveraging spectroscopic calibration to mitigate redshift uncertainties and intrinsic alignment systematics, thereby enabling direct calibration of dominant errors with instruments like DESI.

The Gist

Here is an explanation of the paper "Steeling Weak Lensing Source Galaxy Samples against Systematics using Wide Field Spectroscopy," translated into simple, everyday language with creative analogies.

The Big Picture: Weighing the Invisible Universe

Imagine the universe is a giant, invisible ocean made of Dark Matter. We can't see the water, but we can see how it bends the light from distant stars and galaxies, much like how a funhouse mirror distorts your reflection. This bending is called Weak Lensing.

Astronomers want to use this "cosmic funhouse mirror" to weigh the universe and figure out how fast it's expanding. To do this, they need to look at millions of faint, distant galaxies. The more galaxies they look at, the better their picture should be, right?

The Problem:
The current plan (the "Gold" sample) is to look at every faint galaxy they can find. But there's a catch:

1. The Glasses are Foggy: We don't know exactly how far away these galaxies are (their "redshift"). It's like trying to guess the distance of a car at night just by how bright its headlights look. If you guess wrong, your map of the universe is wrong.
2. The Mirrors are Bent: Galaxies aren't perfect spheres; they are shaped like footballs. Sometimes, they naturally line up with each other due to gravity, not because of the lensing effect. This is called Intrinsic Alignment. It's like trying to measure the wind by looking at leaves, but some leaves are glued together.
3. The Water is Turbulent: The gas and stars inside galaxies (baryons) push back against gravity, messing up the smooth distribution of dark matter on small scales. It's like trying to measure the ocean's depth while standing in a storm.

Because of these "systematics" (errors), looking at more galaxies (the Gold sample) doesn't necessarily give a better answer. The fog and the bent mirrors drown out the extra data.

The Solution: The "Steel" Sample

The authors propose a new strategy: The Steel Sample.

Instead of trying to measure every faint galaxy, they suggest picking a smaller, brighter, and easier-to-measure group of galaxies. Think of it like this:

• The Gold Sample: A massive crowd of people in a dark stadium. You can see everyone, but you can't tell who is who, and you can't hear them clearly.
• The Steel Sample: A smaller group of people in the front row, holding bright flashlights. You can't see as many people, but you can see exactly who they are and where they are standing.

How It Works: The "TrueDirCal" Strategy

The paper suggests using a powerful telescope called DESI (which acts like a high-speed ID scanner) to take "spectroscopic" measurements of a specific subset of these galaxies.

1. The Parent Crowd: Start with a huge list of galaxies from a camera (LSST).
2. The Filter: Pick out the ones that are bright enough and in a color range where the DESI scanner can get a perfect ID (redshift) 95% of the time.
3. The Calibration: Use the DESI scanner to measure the exact distance of these "Steel" galaxies. Because you have the exact distances, you can build a perfect map of where the "parent" crowd is, even the ones you didn't scan.

The Analogy:
Imagine you want to know the average height of everyone in a massive city.

• Old Way (Gold): You try to guess the height of 10 million people by looking at them from a helicopter. You get a lot of data, but your guesses are all over the place because of the distance and angle.
• New Way (Steel): You pick 50,000 people who are standing in a specific, well-lit park. You walk up to them with a tape measure (DESI) and measure them exactly. Because you know exactly how tall these people are, you can mathematically figure out the average height of the whole city with much higher precision, even though you measured fewer people.

Why This is a Game Changer

The paper runs complex computer simulations (forecasts) to see how well this works. Here are the three big discoveries:

1. Less is More (The Saturation Effect)
The authors found that once you account for the "turbulent water" (baryonic physics), looking at more galaxies doesn't help much. The information "saturates."

• Analogy: Imagine trying to hear a song in a noisy room. If you add 100 more microphones, you just hear more noise. But if you move the microphones closer to the singer (better calibration), you hear the song clearly. The Steel sample is about moving the microphones closer, not adding more of them.
• Result: A sample with only 5 galaxies per square arcminute (Steel) is just as powerful as a sample with 27 galaxies (Gold), provided you know their distances perfectly.

2. Self-Correction (The Magic Mirror)
The "Steel" galaxies are so well-separated in distance that the "Intrinsic Alignment" error (the glued leaves) can be calculated and removed automatically by the math itself.

• Analogy: If you have a group of people standing in a straight line, and you know exactly where they are, you can mathematically figure out how much the wind is blowing them sideways without needing a separate wind sensor. The Steel sample is so "clean" that it calibrates its own errors.

3. The Redshift Requirement
To make this work, you need to know the distance to these galaxies with extreme precision (within 0.5%).

• Analogy: If you are aiming a laser pointer at a target 10 miles away, a tiny wobble in your hand (redshift error) makes you miss the target. The Steel sample requires a very steady hand, but because the target is "brighter" (easier to measure), we can achieve that stability with current technology (DESI).

The Conclusion

The paper argues that the future of cosmology shouldn't be about collecting more data blindly. Instead, we should be smarter about the data we collect.

By switching from a "Gold" strategy (massive, fuzzy, hard-to-calibrate data) to a "Steel" strategy (smaller, brighter, perfectly calibrated data), we can:

• Avoid the biggest sources of error.
• Get more accurate answers about Dark Energy and Dark Matter.
• Use existing tools (like DESI) to do the heavy lifting of calibration.

In short: Don't try to count every grain of sand on the beach to understand the ocean. Pick a handful of perfect, clean grains, measure them exactly, and use that to understand the whole beach. That is the power of the Steel Sample.

Technical Summary

Here is a detailed technical summary of the paper "Steeling Weak Lensing Source Galaxy Samples against Systematics using Wide Field Spectroscopy."

1. Problem Statement

The paper addresses the critical systematic uncertainties hindering Stage IV weak lensing surveys (such as LSST/Rubin and Euclid). While these surveys aim to measure the growth of structure and dark energy via 3$\times$2-point analyses (combining galaxy clustering, galaxy-galaxy lensing, and cosmic shear), their precision is threatened by three main factors:

1. Redshift Calibration: Current photometric redshift ($z_p$) calibration methods struggle to achieve the required precision ($\sigma(\langle z \rangle) \sim 0.001(1+z)$) for faint source galaxies due to incomplete spectroscopic reference catalogs and complex selection functions.
2. Intrinsic Alignments (IAs): The alignment of galaxy shapes with the local matter distribution mimics the lensing signal. Current models often assume simplistic redshift evolution, which can bias cosmological parameters if the true evolution is more complex.
3. Baryonic Physics: Feedback from active galactic nuclei and supernovae alters the matter power spectrum on small scales ($k > 1 h \text{ Mpc}^{-1}$), introducing significant uncertainty in theoretical modeling.

The prevailing strategy has been to use extremely dense galaxy samples (e.g., the LSST "Gold" sample with $\bar{n} \approx 27.7 \text{ arcmin}^{-2}$) to overcome statistical noise. However, the authors argue that this approach exacerbates systematic errors because the required spectroscopic calibration for such faint, dense samples is currently infeasible.

2. Methodology

The authors propose a new strategy: the "Steel" sample. Instead of maximizing galaxy density, they propose selecting a sparser, brighter subset of source galaxies that can be directly calibrated using wide-field spectroscopy (e.g., DESI).

• Data Assumptions:

  • Lenses: DESI Bright Galaxy Sample (BGS) and Luminous Red Galaxies (LRGs) over the LSST area ($14,300 \text{ deg}^2$). These are assumed to have negligible redshift uncertainty due to high spectroscopic completeness.
  • Sources (Steel): A magnitude-limited sample ($i < 23.5 - 24.0$) with a target number density of $\bar{n} \approx 5 \text{ arcmin}^{-2}$.
  • Comparison: The "Gold" sample (LSST Y10, $\bar{n} \approx 27.7 \text{ arcmin}^{-2}$) is used as a benchmark, assuming both pessimistic and optimistic redshift calibration scenarios.

• Theoretical Framework:

  • 3$\times$2-point Analysis: Combines angular power spectra of galaxy clustering, galaxy-galaxy lensing, and cosmic shear.
  • Modeling: Uses Hybrid Effective Field Theory (HEFT) and Lagrangian Perturbation Theory for galaxy bias and matter power spectra.
  • Intrinsic Alignments: Adopts a flexible, second-order Lagrangian expansion for IA, allowing the amplitude to evolve as a spline in redshift for each source bin. This avoids the biases of rigid power-law assumptions.
  • Baryonic Effects: Marginalizes over baryonic feedback uncertainties using lensing counter-terms (expanding the power spectrum at high $k$) rather than applying hard scale cuts, preserving information on quasi-linear scales.

• Forecasting:

  • Fisher matrix forecasts are performed to estimate constraints on $S_8$ (a proxy for $\sigma_8$ and $\Omega_m$).
  • Systematic uncertainties (redshift shifts, widths, outlier fractions, IA amplitudes) are treated as free parameters with flexible priors.

3. Key Contributions

1. The "Steel" Sample Concept: The paper introduces a paradigm shift from "dense but poorly calibrated" to "sparse but perfectly calibrated." By selecting brighter galaxies ($i < 23.5-24.0$), the sample becomes accessible to current spectroscopic instruments like DESI, enabling TrueDirCal (True Direct Calibration) where the redshift distribution $n(z)$ is measured directly from the target population.
2. Saturation of Information Content: The authors demonstrate that due to uncertainties in baryonic physics, the cosmological information in weak lensing saturates on quasi-linear scales. Therefore, increasing source density beyond $\sim 5 \text{ arcmin}^{-2}$ yields diminishing returns unless baryonic physics is understood to the $\sim 0.5\%$ level.
3. Self-Calibration of IAs: The study shows that 3$\times$2-point analyses with DESI lenses can self-calibrate complex IA redshift evolution models. The overlap between narrow lens and source distributions breaks the degeneracy between IA and cosmological parameters, making external IA priors less critical for the Steel sample compared to the Gold sample.
4. Feasibility Study: A preliminary assessment using HSC data suggests that a Steel sample with $\bar{n} \approx 5 \text{ arcmin}^{-2}$ can be selected with a spectroscopic success rate $>95\%$, requiring only $\sim 40,000$ spectra (achievable with DESI) to calibrate the $n(z)$ to the required precision.

4. Key Results

• Constraining Power: A Steel sample with $\bar{n} = 5 \text{ arcmin}^{-2}$ and redshift calibration at $\sigma(\langle z \rangle) = 0.005$ provides tighter constraints on $S_8$ than the much denser Gold sample ($\bar{n} = 27.7 \text{ arcmin}^{-2}$) if the Gold sample is calibrated only to Stage-III levels ($\sigma(\langle z \rangle) \approx 0.01(1+z)$).
• Redshift Requirements: The constraining power of the Steel sample saturates when the uncertainty on the mean redshift of the core distribution reaches $\sigma(\Delta z_{core}) \approx 0.005$. Further improvements in density or redshift precision yield minimal gains unless baryonic uncertainties are reduced.
• IA Modeling:
  • The Steel sample is robust against mis-modeling of IA redshift evolution due to its narrow $n(z)$ bins.
  • Overly simplistic IA models (e.g., constant amplitude per bin) can bias $S_8$ by $>1\sigma$ for the Gold sample, whereas the Steel sample remains largely unbiased even with flexible IA models.
  • 2$\times$2-point analyses (clustering + lensing) are sufficient to self-calibrate IA amplitudes for the Steel sample.
• Baryonic Feedback: Baryonic effects are the dominant limiting factor. If baryonic feedback is only known to the $\sim 5\%$ level (current state), the Steel sample outperforms the Gold sample. If baryonic physics were understood to $0.5%$, the Gold sample's higher density would become advantageous again.
• Observational Feasibility: Using HSC data, the authors show that selecting galaxies with $i < 23.5$ allows for a "localizable fraction" (where DESI can secure redshifts) of $\sim 50\%$, resulting in a viable Steel sample density of $\sim 5 \text{ arcmin}^{-2}$.

5. Significance

This work fundamentally challenges the assumption that "more galaxies = better cosmology" in the era of Stage IV surveys. It argues that systematic control via spectroscopic calibration is more valuable than raw statistical power from dense, uncalibrated samples.

• Strategic Shift: It advocates for a "TrueDirCal" approach where a spectroscopic instrument (like DESI) is used not just for lenses, but to directly calibrate the source population of a weak lensing survey.
• Cost-Effectiveness: The Steel sample requires significantly fewer spectroscopic resources to achieve high-precision calibration compared to the Gold sample, making it a more feasible path to unbiased cosmological constraints.
• Robustness: By utilizing flexible theoretical models (HEFT, spline IAs) and focusing on quasi-linear scales, the Steel sample mitigates the dominant systematics (redshift, IA, baryons) that currently plague weak lensing analyses.

In conclusion, the paper posits that a sparser, spectroscopically calibrated "Steel" sample is the optimal strategy for next-generation weak lensing surveys, offering superior cosmological constraints by prioritizing systematic mitigation over statistical density.