The Hubble sequence in JWST CEERS from unbiased galaxy morphologies

By constructing resolution-matched "absolute" images of galaxies from HST and JWST surveys to ensure consistent structural analysis, this study reveals that a continuous Hubble-like morphological sequence is already established by redshift z~4, with massive galaxies evolving via either stable disk growth or rapid compaction-driven quenching from irregular progenitors.

The Gist

The Big Question: Do Galaxies Grow Up?

Imagine you are looking at a family photo album. You see babies, toddlers, teenagers, and adults. You know that a baby looks different from an adult. But what if you could see a baby and an adult standing next to each other, and they looked exactly the same? Would you believe they were related?

That is the puzzle astronomers are trying to solve with galaxies. For decades, we've had a "family tree" for galaxies called the Hubble Sequence. It's like a sorting system:

• Late-Type (LTGs): These are the "spiral" galaxies (like our Milky Way). They are blue, full of gas, and busy making new stars. Think of them as the energetic, messy teenagers.
• Early-Type (ETGs): These are the "elliptical" galaxies. They are red, smooth, and have stopped making stars. Think of them as the calm, retired grandparents.

The big question is: Did this sorting system exist when the universe was young? Or did galaxies start as messy blobs and slowly organize themselves into this neat pattern over billions of years?

The Problem: The "Zoom" and "Brightness" Trap

The problem is that looking at the universe is like trying to compare a high-definition photo of a person standing 1 meter away with a blurry, dark photo of a person standing 10 kilometers away.

1. Distance makes things look different: A galaxy far away looks smaller and dimmer.
2. Time travel is tricky: When we look at a distant galaxy, we are seeing it as it was billions of years ago.
3. The Bias: If you try to compare a nearby galaxy to a faraway one using standard telescopes, the faraway one looks "messier" just because the image is blurry and faint. It's hard to tell if it's actually a messy galaxy or just a clear galaxy that looks messy because it's far away.

The Solution: The "Universal Zoom"

To fix this, the authors (Elizaveta Sazonova and her team) invented a clever trick. They took thousands of galaxy images from the Hubble Space Telescope (for nearby ones) and the James Webb Space Telescope (JWST, for faraway ones) and processed them to look as if every single galaxy was sitting at the exact same distance from us: 10 million light-years away.

Think of it like this:

• Imagine you have photos of a dog, a cat, and a hamster taken from different distances.
• The team used a computer to resize, brighten, and sharpen every photo so that every animal looks like it's sitting on your lap.
• They also added a little bit of "noise" (static) to the close-up photos so they looked just as grainy as the distant ones.

Now, they can compare the "dog" (nearby galaxy) and the "hamster" (distant galaxy) fairly. If the hamster looks like a hamster even when it's "close up," then it really is a hamster, not a blurry dog.

What They Found: The "Hubble Sequence" is Old

After creating these "absolute" images, they used a smart computer algorithm (called UMAP) to map out the shapes of 2,825 galaxies. Imagine a giant playground where every galaxy is a kid. The computer grouped them based on how they looked.

The Result:
They found that the "Hubble Sequence" (the sorting of galaxies into spirals and ellipses) already existed when the universe was only 2 billion years old (about 12 billion years ago).

• No "Growing Pains": There wasn't a long period where everything looked like a messy blob. The "teenagers" (spiral galaxies) and the "grandparents" (elliptical galaxies) were already distinct from each other way back then.
• The "Bat" Shape: When they plotted the galaxies on a graph, they formed a continuous curve that looked a bit like a bat. The left side was the spirals, the right side was the ellipses, and the middle was a smooth transition. There were no gaps.

The Two Paths of Massive Galaxies

The team also looked at how these galaxies changed over time by tracing their "family trees" (using their mass to guess who their ancestors were). They found two very different stories for massive galaxies:

1. The Stable Path (The "Forever Disk"):
   Some massive galaxies started as spinning disks and stayed that way. They kept making stars and growing, but they never changed their shape. They are like a person who stays fit and active their whole life.

2. The "Compaction" Path (The "Blue to Red" Switch):
   Other massive galaxies started as messy, irregular blobs (like a pile of laundry). Then, something happened—maybe they crashed into each other or their gas collapsed inward. They quickly shrank down, became very dense, stopped making stars, and turned into smooth, red elliptical galaxies.

   • The Analogy: Imagine a chaotic party (the irregular galaxy). Suddenly, everyone rushes to the center of the room, the music stops, and the room becomes a quiet, orderly library (the elliptical galaxy). This happened very fast (in just a few billion years).

The "Irregular" Mystery

They also found a group of "irregular" galaxies that didn't fit the neat spiral or ellipse categories. These were the "troublemakers" of the universe.

• At high redshifts (very far back in time), there were lots of these messy galaxies.
• As time went on, the number of messy galaxies dropped, and the number of neat elliptical galaxies rose.
• Conclusion: It seems the "messy" galaxies were the ancestors of the "neat" elliptical ones. They were the raw material that got compacted and cleaned up.

Why This Matters

This paper solves a long-standing debate. Some scientists thought the universe was just a chaotic mess in the beginning and only organized itself later. This study says: "No, the organization was there from the start."

By using the "Universal Zoom" technique, they proved that the rules of galaxy shapes haven't changed much over the last 12 billion years. The universe has been sorting its galaxies into spirals and ellipses for almost its entire history.

In short: The universe isn't a messy construction site that finally got organized; it's a well-organized city that has been building its neighborhoods in the same way for billions of years.

Technical Summary

1. Problem Statement

The existence and timing of the "Hubble sequence" (the morphological classification of galaxies ranging from late-type disks to early-type ellipticals) at high redshifts ($z \sim 4$) remains a subject of debate. Previous studies using HST and early JWST data have produced conflicting results:

• Visual classifications suggest a Hubble sequence persists up to $z \sim 5$.
• Quantitative studies often find that high-redshift "early-type" galaxies lack the central concentration (bulges) seen in local counterparts, challenging their classification.
• The Core Challenge: Comparing galaxy morphologies across cosmic time is biased by observational effects. As redshift increases, galaxies appear smaller (angular size), dimmer (cosmological surface brightness dimming), and are observed at different rest-frame wavelengths (morphological k-correction). These factors alter structural parameters (e.g., Sérsic index, asymmetry), making it difficult to distinguish intrinsic evolution from observational artifacts.

2. Methodology

To address these biases, the authors constructed "absolute" images of galaxies, effectively simulating how they would appear if observed from a fixed distance of 10 Mpc.

Data Selection:

• Surveys: Combined data from JWST CEERS and HST CANDELS.
• Sample: 2,825 galaxies spanning $0.15 < z < 4.5$ and stellar masses $8 < \log M_\star < 11$.
• Filtering: Filters were selected to trace the same rest-frame optical wavelength (avoiding the D4000 break) to ensure consistent stellar population tracing across redshifts.

Image Processing ("Absolute" Imaging):
The authors applied a rigorous pipeline to match effective resolution and surface brightness limits:

1. Resolution Matching: Images were convolved with a Gaussian PSF and resampled so that the Petrosian radius ($R_p$) always spans 13.6 pixels (approx. 4 PSF elements per radius). This accounts for intrinsic size evolution and mass-size relations.
2. Depth Matching: Flux was adjusted for cosmological dimming and mass-to-light ratio evolution. Gaussian white noise was added to all images to match a uniform $1\sigma$ surface brightness limit of 23.7 mag/arcsec².
3. Segmentation: A "hot-cold" algorithm masked foreground/background contaminants before redshifting to prevent merger-like artifacts in the analysis.

Morphological Analysis:

• Feature Extraction: 59 structural parameters were calculated using statmorph-lsst (including Gini, $M_{20}$, asymmetry, concentration, and isophotal ratios).
• Dimensionality Reduction: The authors used UMAP (Uniform Manifold Approximation and Projection) to reduce the 24 selected parameters into a 2D morphological phase space ($U_1, U_2$). Unlike PCA, UMAP captures non-linear relationships, allowing for the visualization of continuous morphological sequences.
• Progenitor Tracking: Using empirical mass assembly histories, galaxies were binned by their expected present-day mass ($\log M_{\star,0}$) to trace the evolution of specific progenitor populations.

3. Key Contributions

• Bias-Corrected Morphology: The creation of a dataset where galaxies are compared at identical effective resolution and depth, removing the primary source of systematic error in high-redshift morphology studies.
• UMAP Phase Space: The application of UMAP to galaxy morphology reveals a continuous, non-bimodal distribution rather than distinct clusters, providing a more nuanced view of the Hubble sequence.
• Quantitative vs. Visual Comparison: A direct comparison showing that the authors' parametric approach yields a cleaner separation between Late-Type (LTG) and Early-Type (ETG) galaxies than visual classifications or standard deep learning models trained on visual data.

4. Key Results

A. The Existence of the Hubble Sequence

• The UMAP phase space reveals a continuous sequence from late-type to early-type galaxies with no redshift gradient.
• A "mass-morphology main sequence" exists up to $z \sim 4.5$ (lookback time $\sim 12.5$ Gyr).
• Conclusion: The Hubble sequence is established early; LTGs and ETGs at $z \sim 4$ are morphologically indistinguishable from their local counterparts when observed at matched resolution.

B. Classification Performance

• The UMAP-based classification separates LTGs and ETGs more effectively than visual classifications (e.g., Ferreira et al. 2023), which show significant overlap in Sérsic indices between "disks" and "spheroids."
• The method performs comparably to deep learning (Huertas-Company et al. 2024) but avoids the biases inherent in training sets based on visual labels.

C. Evolution of Low-Mass Progenitors ($\log M_{\star,0} \in [9, 10]$)

• Stability: Progenitors of low-mass galaxies are predominantly star-forming disks (LTGs) at all epochs.
• Lack of Evolution: Their morphology does not change significantly over time. They remain star-forming, with only a slight increase in quenching at low redshifts due to environmental effects.
• Caveat: Due to resolution limits, substructures (spiral arms, bars) are not visible in high-$z$ LTGs, making them appear "featureless" compared to local galaxies.

D. Evolution of Massive Progenitors ($\log M_{\star,0} \in [10, 11]$)

• Two Distinct Paths:
  1. Stable Disks: A population of LTGs that remains star-forming and disk-like throughout cosmic time.
  2. Compaction/Quenching: A population that evolves from irregular progenitors into ETGs.
• Timeline: At high redshift ($z > 2$), massive progenitors are dominated by irregular galaxies ($\sim 50\%$). Over a few Gyr, these irregulars undergo rapid compaction, form central spheroids, and quench star formation, becoming ETGs.
• Mechanism: This supports a "blue nugget" $\to$ "red nugget" scenario, likely triggered by mergers or disk instabilities.

E. The "Irregular" Population

• Galaxies outside the main sequence in UMAP space are heterogeneous.
• Clustering analysis identifies distinct sub-groups corresponding to different merger stages (e.g., double nuclei, tidal tails) and low-surface-brightness systems.
• The fraction of irregulars decreases as the ETG fraction increases, suggesting a direct evolutionary link.

5. Significance

This paper resolves a major tension in extragalactic astronomy by demonstrating that the Hubble sequence is a robust feature of the universe from $z \sim 4.5$ to the present.

• Methodological Impact: It establishes that previous contradictions between visual and quantitative studies were largely due to uncorrected observational biases (resolution and surface brightness).
• Evolutionary Insight: It clarifies that galaxy evolution is not a single path. Low-mass galaxies are stable disks, while massive galaxies follow a bifurcated path: some remain disks, while others undergo rapid compaction and quenching from irregular states.
• Future Directions: The study highlights the need for deeper surveys (e.g., JADES) to resolve the nature of irregular progenitors and the potential role of AGN in mimicking compact early-type structures at high redshift.