The sky is brighter than astronomers imagined

Although we’ve now firmly entered the JWST era in astronomy, our deepest views of the faintest objects of all still come courtesy of the Hubble Space Telescope. Now working in its 35th year since launch, Hubble has spent more time viewing certain specific, dedicated regions of sky than any observatory ever, achieving unprecedented depths in the process. The deepest views of the Universe ever remain the Hubble Ultra Deep Field, with a cumulative total of 11 days of observing time (across all wavelengths of light), and a portion of the Ultra Deep Field that was cumulatively imaged for roughly twice as long: the Hubble eXtreme Deep Field.

Even though JWST has also viewed this region, the shorter observing times mean that, in many ways, Hubble’s views are more sensitive to fainter signals, particularly at optical (visible light) wavelengths. However, there’s something that’s a little bit peculiar about the bright, massive galaxies that appear in these Hubble images: it’s as though their light abruptly stops at a particular edge for each galaxy, rather than diffusely continuing the way we see light continuing within most nearby, extended galactic objects. A 2018 reanalysis of the raw data from these fields showed that the extended, diffuse light at the outskirts of these galaxies is real, and that the methods use to produce the more-famous images commonly shown are artificially removing starlight that’s very, very real. Here’s what everyone should know about it.

The Hubble eXtreme Deep Field (XDF) may have observed a region of sky just 1/32,000,000th of the total, but was able to uncover a whopping 5,500 galaxies within it: an estimated 10% of the total number of galaxies actually contained in this pencil-beam-style slice. The remaining 90% of galaxies are either too faint or too red or too obscured for Hubble to reveal, but when we extrapolate over the entire observable Universe, we expect to obtain a total of ~2 trillion galaxies.

Above, you can see the Hubble Ultra Deep Field, with the Hubble eXtreme Deep Field inlayed within it, in all their glory. These views represent some of the deepest images of the ultra-distant Universe ever taken, where many of the galaxies shown here are billions or even tens of billions of light-years away, with the most distant of all being around ~32 billion light-years distant. This is due to the combinations of facts that:

• it’s been 13.8 billion years since the start of the hot Big Bang,
• the Universe is expanding,
• and has been expanding the entire time,
• including by expanding faster in the past than it is today,
• and that the light that we observe, right now, is the light that is only arriving at this moment,
• after journeying through that expanding Universe from the moment it was emitted by that distant galaxy until its arrival.

The youngest galaxies in this field are the farthest away, and the ~32 billion light-year distance corresponds to light that’s been traveling for about 13.4 billion years: emitted when the Universe was much younger and less evolved.

These galaxies, whether they’re nearby, at intermediate distances, or extremely far away, all seem to have one property in common: they appear sharp, in focus, and just beyond the extent of their luminous regions, appear to fade directly into the black abyss of deep, empty space. But is this truly reflective of how these galaxies (and the background of these galaxies) really appear? Or is this simply an artifact of how these images were created?

Shown here are two views of the nearby, Milky Way-like galaxy Messier 63: the Sunflower galaxy. While the stellar extent of the galaxy is normally only shown to overlap with the bright dusty disk, superior techniques can reveal the low surface-brightness halo and streams around it. Those extended features show the true stellar extent of the galaxy, and is required to make an accurate determination of its total stellar mass.

To start with, let’s take a look at the properties of nearby galaxies. For example, you can see, above, a galaxy commonly known as the Sunflower galaxy: Messier 63, or NGC 5055. This is a nearby spiral galaxy similar to the Milky Way in many ways, like size, but without a central bar and with only loosely-defined spiral arms. You’ll notice, however, that this galaxy has been animated to “flash” between an optical view of the main stellar extent of the disk of the galaxy, alone, and an inverted, extended view of the galaxy that showcases the full extent of the stellar halo and stellar streams surrounding it. As you can clearly see, even visually, there are a lot more stars to this galaxy than you’d infer simply by looking at the stellar disk alone.

And that’s common: it turns out that many, probably most, and maybe even all large, massive spiral and elliptical galaxies have:

• large, extended stellar halos that go well beyond the extent of the main, bright galactic disk,
• streams of stars that surround them, likely from galactic interactions and the cannibalism of smaller galaxies,
• as well as an ellipsoidally-configured bulge of stars that diffusely continues on in all three dimensions past the bright edges of the main galaxy.

In other words, what we see for nearby galaxies, at short cosmic distances, should apply as well to more distant galaxies at great cosmic distances.

The blank region of sky, shown in the yellow L-shaped box, was the region chosen to be the observing location of the original Hubble Deep Field image. With no known stars or galaxies within it, in a region devoid of gas, dust, or known matter of any type, this was the ideal location to stare into the abyss of the empty Universe. Today, we know of even more pristine regions than we did in the early 1990s.

So let’s ask a key question, then: when we acquire these images of the distant Universe, such as with the Hubble Space Telescope, how do we do it?

The starting point is with a “blank” patch of sky, like you see up above. This was, in fact, the patch of sky that was chosen to be the target of the original Hubble Deep Field back in the mid-1990s. This patch of sky was chosen for a number of important properties that it possessed.

• This region of sky was well out of the galactic plane, away from the gas, dust, and stars that would interfere with deep observations.
• This region of sky had no known bright stars or nebulae within it; the only stars within the Milky Way were faint, cool, and distant: the red dwarf stars that are difficult to find even from up close.
• And there were no major extragalactic objects, like galaxies or galaxy clusters, located anywhere within this field-of-view.

It was pristine, and that meant there were no bright light sources (or light-blocking, dusty sources) within this region of sky at all. If you want to expose a region of sky and collect large numbers of photons, looking at a “quiet” patch of sky like this is the place you ought to go to. These criteria were applied to all subsequent deep field images taken with flagship telescopes like Hubble.

This photo shows the Hubble Space telescope being deployed, on April 25, 1990, one day after its launch. It was taken by the IMAX Cargo Bay Camera (ICBC) mounted aboard the space shuttle Discovery. Hubble was humanity’s first flagship-class space telescope, and remains operating even today: 35 years after its initial deployment.

However, you can’t simply acquire all of the data, over time, and add up all of the registered photons that show up in your detector. If you do that, you’ll wind up with all sorts of artifacts in your image that aren’t real.

• There are cosmic rays that strike the detectors randomly, creating “hot pixels” that make sources of light that correspond to no actual sources.
• There’s stray light that enters the telescope, often after multiple reflections, and that can pollute what your instruments register.
• There’s light generated from thermal noise, where the fact that the telescope and its instruments is not extremely cold means that additional signals (again, that don’t correspond to real objects) get registered in the detector.
• And there’s background light from all over the sky from faint, unresolved sources that nevertheless will contribute to the intensity of pixels in your final image.

There are other sources of noise as well: asteroids or other Solar System objects that transit through a frame (or a few frames) of exposure, for instance.

The way we combat these, typically, is with image reduction techniques that are usually applied before an observer using the telescope data ever sees it. These techniques include flat fielding, which helps calibrate all digital images, subtraction of after-effects, which helps remove the errors that result from hot pixels, cosmic rays, or transiting asteroids or satellites that leave “streaks” or persistent afterimages in the detector, background sky subtraction, which helps remove the background light from unresolved sources, and additional effects from the fact that different wavelength filters must be added together, even though those images will have different resolutions and detector responses.

The original Hubble Deep Field image, for the first time, revealed some of the faintest, most distant galaxies ever seen. Only with a multiwavelength, long-exposure view of the ultra-distant Universe could we hope to reveal these never-before-seen objects.

When you take a look at an image like you see above — which is the original Hubble Deep Field image in this case — you’re seeing it with all of these subtraction techniques already applied. So do the scientists who use these data for scientific purposes, including:

• counting galaxies,
• estimating the mass and angular size of these galaxies,
• calculating the amount of stellar mass within this galaxies,
• and drawing conclusions about galaxy evolution over cosmic time, based on these images.

As long as all of this is done, well, perfectly, we won’t have anything to worry about. We’ll be counting all of the galaxies down to the limits of what our telescopes can expose. We’ll estimate a total mass, angular size, and stellar mass for each these galaxies that accurately reflects what the telescope can see in the observing time that was dedicated to looking at these galaxies. And, if this is done equally well for every galaxy in these deep field images, we can draw accurate, responsible, robust conclusions about galaxy evolution based on what we’re seeing from examining the imaged objects.

But there’s a big danger here. It’s possible that some of the image reduction techniques we’ve been using are overly aggressive: highlighting only the areas very rich in stars (or above a certain brightness) within each galaxy, and overly subtracting the diffuse, extended light coming from beyond the main stellar disk or halo of a galaxy.

This mosaic, released in 2018, is a portion of the GOODS-North field as imaged by Hubble over a total observing time of around 11 days: the Hubble Ultra Deep Field. Even in the JWST era, this is perhaps the deepest image of the Universe ever acquired for an area so large on the sky.

It was only relatively recently, in 2018, that a team of researchers decided to go back to basics for the various Hubble Deep Fields, and to create a new image reduction pipeline for these views of the deep, ultra-distant Universe. The pipeline they created was known as ABYSS, which was specifically designed to optimize the ability to image the low surface-brightness Universe: exactly what we need if we want to examine the extended stellar halos of distant and ultra-distant galaxies.

Above, you can see the standard “Hubble Ultra Deep Field” on its own, which looks like it only highlights high surface-brightness objects, or high surface-brightness features within these objects. I’ve downloaded this image and played with the contrast and brightness settings for myself, and even by fiddling with those parameters to the extreme, I’m unable to bring out any additional detail on the outskirts of even the brightest, largest (in terms of angular scales) galaxies that are in this image.

But in developing ABYSS, where those techniques should also be applicable to ESA’s Euclid, NASA’s JWST, and future observatories like the Nancy Roman Space Telescope, the researchers sought to create a more robust, more careful, and less aggressive background subtraction method.

This figure, taken directly from the published paper on the ABYSS pipeline as applied to the Hubble Ultra Deep Field, presents a very different view of the galaxies within it: one that’s sensitive to the low surface-brightness features present in the extended halos of these galaxies. Those features are otherwise invisible in the originally released Ultra Deep Field images.

When they looked at the same region of sky, with the same raw (uncalibrated) data but subjecting it to their own calibration pipeline, they got a very different view. In particular, they found that there was an enormous amount of low surface-brightness structure in practically all of the galaxies present, especially in the large, massive galaxies that took up the greatest angular areas on the sky. The researchers noted that there were four specific points in the data reduction that offered improvements over the de facto techniques used by astronomers processing Hubble data for nearly three decades:

• they created new absolute sky flat fields,
• they improved the extended persistence models that affect sky correction,
• they developed a new end-to-end algorithm for performing dedicated sky background subtraction, allowing for better matching between background levels before co-adding the various frames together,
• and they identified and removed a number of sources of bias that affect the final signal-to-noise ratios of galaxies due to co-adding all the relevant frames together.

One thing that didn’t change from the original Ultra Deep Field to the re-analyzed Ultra Deep Field was the number of galaxies that were found; they were identical. However, the features of each individual galaxy, particularly for the larger galaxies in terms of angular size, showed differences that were quite astonishing.

Whereas the “standard” version of the Hubble Ultra Deep Field shows the bright, central regions of galaxies and only the black depths of space between them, a more accurate picture reveals the extended galactic halos and the brightness of the stars within them. Only the white regions are truly dark in this image; everywhere you see black represents the presence of an extended halo of stars surrounding the relevant galaxy.

Here, you can see a zoomed-in view of a few hundred of the galaxies found in the Ultra Deep Field but with the ABYSS data reduction pipeline (instead of the standard pipeline) applied. As you can clearly see, there’s a “black” halo-like region around each galaxy. That is not an image artifact; those are the stellar halos, similar to the stellar halo we imaged earlier for nearby galaxy Messier 63, for these ultra-distant galaxies.

This is a big deal!

When you are looking at published data from the Hubble Space Telescope — and arguably, from any telescope, including JWST, Euclid, or even a ground-based telescope — you have to recognize that whether the telescope (and the data reduction pipeline) has exposed these stellar halos or not, they’re very real and they play a significant role in terms of astrophysics. They teach us, when we image them properly, that:

• galaxies are physically larger, in the extent of their stars,
• galaxies are brighter, overall,
• galaxies are more massive, in terms of total mass, including dark matter,
• and galaxies have larger stellar masses inside of them,

than most images, even Hubble images, reflect them to be. It’s absolutely necessary to make the most of your data, and unless you calibrate your images correctly, you run the risk of throwing good information away, and drawing conclusions that aren’t backed by the actual data as a result.

This image shows the full depth of the mosaic of the ABYSS version of the Hubble Ultra Deep Field, taken in infrared light with Hubble’s Wide Field Camera 3. The extended galactic halos are revealed in black, surrounding the main stellar extent of the galaxies.

All told, across the entire Hubble Ultra Deep Field, there are between 100 and 200 billion extra “stellar masses” worth of stars when you add back in this extra light that had been oversubtracted from the originally released images. This corresponds to somewhere between two and three times the total amount of stellar mass present in the Milky Way, which corresponds to several hundred billion, and perhaps even a trillion or more, stars altogether. While it didn’t reveal any “missing” galaxies using this technique, it certainly does change the story of galactic evolution that these images, and similar images like them, tells us about the Universe.

The key advance is that we now know how to reveal the extended, low surface-brightness outskirts of even the faintest extragalactic sources whenever we conduct and ultra deep galaxy survey. We can, as the astronomy community, apply these improved techniques to not only current and future deep field observations, but we can go back and do this retroactively for any Hubble (or other) images that may have suffered from an accidental oversubtraction of this diffuse light. We’ve long been asking questions about stars, dark matter, the extent of a circumgalactic halo, and much more. With a greater awareness of the presence and power of these new techniques, perhaps the resolution to some of our cosmic mysteries could be on the cusp of coming into focus at last.

Ethan Siegel acknowledges the IAC’s Nacho Trujillo for his clear explanation of the ABYSS pipeline and the astrophysical implications of its application to the Hubble Ultra Deep Field.