Tag Archives: Leah Fulmer

Astronomy behind the scenes: great successes and colossal blunders

The most recent gathering of Astronomy on Tap Seattle promised to take us inside the way science is really done, and delivered with tales of unexpected successes and a colossal fail that left a team of cosmologists with cosmic egg on their faces.

Leah Fulmer

Leah Fulmer, a second year graduate student in astronomy at the University of Washington, gave a talk titled “Falling with Style: How Astronomy’s Most Intriguing Discoveries Happen by Accident.” Fulmer noted that astronomers have lots of choices when it comes to their research. They can select which part of the sky to examine, what to look at, how long to look, how often to look, and in which wavelengths of light to look, just to name a few. There’s lots of potential there.

“Every time we look at the universe in a new way we discover new phenomena that we never even expected to see,” she said. Fulmer shared three historical examples of such scientific serendipity.

The first was the detection of the cosmic microwave background (CMB) back in the 1960s. At the time it was theorized that 400,000 years after the Big Bang the CMB would have left its energy throughout the universe as a result of the event. Arno Penzias and Robert Wilson had access to a big radio telescope and were working on doing some radio astronomy. The problem was that they couldn’t tweak out some pervasive and persistent noise from their observations. Meanwhile down the road some theorists at Princeton were trying to figure out how to detect evidence of the CMB. Penzias and Wilson had already done it!

“By accident they took this telescope that NASA had built for satellite communicaiton, they stuck it out there, and they found literally the origins of the universe,” Fulmer said. “This changed our understanding of astronomy and physics as we know it and it was a really, really big deal, just by looking at something in a new wavelength.”

More recently the operators of the Hubble Space Telescope decided to pick out an empty, black part of the sky and have the scope stare at it for 100 hours. Many scientists thought this was a bit daft.

The Hubble Deep Field. Image credit: Robert Williams and the Hubble Deep Field Team (STScI) and NASA/ESA

“They found what’s now known as the Hubble Deep Field,” Fulmer said. “They found an incredible plethora of galaxies that they never expected to see.” It revolutionized our understanding of the number of galaxies in the universe and added greatly to the types, shapes, and sizes of galaxies that we know about.

The Kepler Space Telescope found thousands of exoplanets, and collected data on so many things that scientists couldn’t possibly look at all of them. They enlisted citizen scientists through Zooniverse to help examine objects. Participants looked at the data and among their findings is an oddly behaving star for which its light curves defy explanation. We now know of it as “Tabby’s Star,” after astronomer Tabetha Boyajian, who wrote the paper about the discovery.

“To this day we don’t actually know what this star is,” Fulmer said. There have been lots of ideas about the odd light curves, from a random pack of asteroids that might be irregularly blocking light, some sort of cosmic catastrophe that kicked up debris, and even giant space structures built by an unknown civilization.

“It’s very precarious for an astronomer to suggest that this might be aliens,” Fulmer laughed, noting that the media would have a field day with that sort of thing.

The potential for discovering strange new things in the universe is about to increase. The Large Synoptic Survey Telescope is scheduled to go online in a few years, and when it does it will collect petabytes of data, doing a complete sweep of the sky every few nights for a decade.

Fulmer said a big part of her job in the project will be to help “develop an algorithm that is going to be able to systematically identify the things that we’ve never seen before.” That’s a tall order, combing all of that data for things we know about, things that have been theorized, and those that come out of the blue.

“We don’t what surprises we might find,” Fulmer said, “but that’s what makes it so exciting.”


Samantha Gilbert, a first-year graduate student in astronomy at the UW, told a story about a colossal and embarrassing failure. Her talk was titled, “Leaving the Competition in the Dust: A CMB Case Study.”

“The story I want to tell you tonight has everything: It has science. It has drama. It has egos. It has really esoteric vector math,” Gilbert said to laughter. “It encapsulates some of the things that are really wrong with how some people do science today.”

The story also involves the cosmic microwave background. Cosmologists are trying to figure out what happened between the Big Bang and the formation of the CMB 400,000 years later. A leading theory is that there was a period of inflation in the moments after the Big Bang during which the universe expanded rapidly. If that happened, it would have created gravitational waves, and those waves would have left behind a pattern in the CMB that we could recognize, called “B-mode polarization.”

A map of the cosmic microwave background. Image credit: NASA / WMAP Science Team

“B-mode polarization is an extraordinarily difficult thing to detect,” Gilbert said, “but proving it exists, proving that inflation really happened by detecting the traces of inflationary gravitational waves” would be Nobel Prize-worthy.

That’s where the intrigue starts. One group striving for this discovery had an experiment called BICEP (Background Imaging of Cosmic Extragalactic Polarization), which was followed by BICEP2, which had more sensitive detectors than the first version and more of them. They found what they were looking for. In fact, the signal of B-mode polarization was even stronger than anticipated. The team declared the discovery during a 2014 news conference at Harvard, issued a video, broke out the bubbly, and in general whipped up lots of hoopla about the discovery.

In the following months some 250 papers were published in response to BICEP2. One of them was from BICEP’s main competitor, the Planck Experiment, and their point was that BICEP’s discovery was bunk and that what they detected was not B-mode polarization, but cosmic dust.

“The fact that BICEP2 had so confidently announced a result that was so quickly disproven had a rippling effect throughout the community,” Gilbert said. “Scientists were horrified because they thought, ‘now the public is going to discredit us, they’re not going to trust us.’ Journalists were also horrified because they felt they had a role in spreading disinformation.”

They were also seeing an ugly side of the scientific community.

The need for speed

How did this happen? BICEP principal investigator Brian Keating wrote a book about their process, titled Losing the Nobel Prize (W.W. Norton & Company, 2018). Gilbert summarized their decision-making.

She said BICEP2 only looked at one wavelength of light so they could get the results as quickly as possible. They knew about the possibility of cosmic dust, but didn’t have the tools to distinguish between dust and B-mode polarization. The Planck folks were thought to have the data, and BICEP asked them to share. They declined.

This led BICEP to jump to the conclusion that Planck also had evidence of B-mode polarization and were aiming to scoop them on the discovery and dash their dreams of a Nobel Prize. So they hurried to make the announcement. This might have worked out OK, if they’d been right, but the BICEP group made one other glaring error.

“They actually hadn’t put their paper through peer review,” Gilbert noted, generating groans among the science-savvy audience at Astronomy on Tap.

“That is a no-no,” she understated. “That is a bad thing to do because peer review is what makes science credible in the first place. It’s a really important check against the dissemination of junk science. You really need other scientists to independently assess your results.”

Gilbert said the bad decisions were all motivated by fear.

“Overly competitive environments are part and parcel of an individualistic conception of science and an individualistic conception of science says that the most important thing is to get a result before your competition,” she said. “When that’s the environment that you’re working in you tend to make decisions based on fear.”

“I would argue that the reason that BICEP2 made these decisions based on fear is that they were operating in such a toxically competitive environment that it became dysfunctional,” Gilbert said. “Whether you think competition is really good for science, really bad, or somewhere in between, I think that this case study shows us that it’s really worth thinking about the ways that we systemically and interpersonally encourage competition, and how that might jeopardize our ways of knowing.”

Gilbert said there’s hope for the future. The hunt for B-mode polarization continues, and BICEP and Planck are teaming up going forward, combining their resources and know-how in the work.

“Competition might be the most efficient way to A result, but collaboration is probably the most efficient way to a RELIABLE result,” she said.

Astronomy on Tap Seattle is organized by graduate students in astronomy at the University of Washington.


More info:

Watch both talks on YouTube

Please support Seattle Astronomy with a subscription through Patreon.

Waffles and big data in the universe

Waffles and big data were on the menu at the most recent gathering of Astronomy on Tap Seattle at Peddler Brewing Company in Ballard.

Leah Fulmer

Fulmer at work. Photo: Astronomy on Tap Seattle

Leah Fulmer, who is working at the University of Washington on her Ph.D. in astronomical data science, gave a talk titled, “Data-Driven Astronomy in the 2020s and Beyond.” Fulmer explained that we’re in the midst of a “data tsunami” that’s been growing over the last three decades of astronomical surveys.

Back in the 1990s the Palomar Digital Sky Survey and the Two Micron All-Sky Survey each collected about a terabyte of data. That’s a trillion bytes; 1012 bytes. Enough to fill a thousand one-gigabyte smartphones.

The 2000s brought the Sloan Digital Sky Survey (SDSS) and the Galaxy Evolution Explorer. These collected in the tens of terabytes of data. In the 2010s Pan-STARRS collected a petabyte of data; a quadrillion bytes.

In the future this astronomical growth in data collection will continue. The Large Synoptic Survey Telescope (LSST) under construction in Chile will survey the entire night sky every few nights for ten years. It will ultimately collect an astounding 500 petabytes of data—that’s 20 terabytes every single night.

“SDSS had a total data collection of 40 terabytes,” Fulmer pointed out. “We’re going to have one SDSS every two nights in the 2020s. This is a big freaking deal.”

On top of the data, Fulmer noted that the LSST will alert its network when it finds something interesting. Given the amount of data, Fulmer said there will be ten million alerts every night, or about 232 every second.

“This is overwhelming; this is a data tsunami,” she said. “With this sort of data collection astronomers cannot do our science in the way we have up until this point.”

A new way to look at data

Up until recently astronomers would apply for telescope time, make their observations, take the data home, and analyze it. That won’t work in the era of big data for a couple of reasons. First, you can’t jam that much data onto your laptop. Second, there just aren’t enough astronomers to sort through data on objects one by one. As you might guess, we need the help of computers.

“Specifically, we need the help of machine learning,” Fulmer said. This can be both “supervised” and “unsupervised” learning. Astronomers can identify objects by their light curves, and the computers can be taught what those are. That’s supervised. In unsupervised learning, the computers can go out on their own and sort various observations into categories with similar characteristics, and we can figure out what’s in each category.

Once you figure that out, a data broker like ANTARES (the Arizona-NOAO Temporal Analysis and Response to Events System, and yes, astronomers still rule at acronyms) can let the right people know about discoveries in a timely manner.

Fulmer said it’s interesting that ANTARES will never look at the sky, just at data, and that many future astronomers may never visit a telescope, just analyze the data. Different fields can learn from each other about how to process all of this information.

Fulmer finds the era of big data exciting.

“It’s not just data-driven astronomy, it’s data-driven everything,” she said.

Astronomy with your breakfast

N. Nicole Sanchez is working on her Ph.D. in astronomy at the UW, and her research interest is in spiral galaxies like our own Milky Way and how they evolve. This, naturally, led her to think of galaxies as waffles. Thus the title of her talk, “Black Holes, Gas, and Waffles.”

Spiral galaxies form into disks, she explained, and a waffle is a disk. The galaxies have a central bulge, represented on the waffle by a big pat of butter. Marshmallows, suspended by toothpicks, represent globular clusters of stars. Red and blue sprinkles represent old red stars and young blue ones. You just have to imagine the supermassive black hole at the center of the waffle. It may be massive, but it’s super small compared to the size of the waffle.

Sanchez came up with the idea for this model while teaching at the UW in the “Protostars” summer science camp for middle school girls the last couple of years. In the waffle model, syrup represents the gas in the galaxy.

“That’s what you’re making your stars out of, so there’s going to be a lot in your disk,” Sanchez said.

In fact, her faculty advisors got wind of the waffle model and said it would need A LOT of syrup, which led to the hilarious twitter thread below. Click on it to see the academic discussion.

Sanchez admitted that her waffle galaxy may be “a bit too simplified” as a model. But the syrup is important.

“There’s actually tons of gas around really all galaxies, in what’s called the circumgalactic medium,” Sanchez said. The gas is important to the evolution of a galaxy. It feeds the black hole and helps  form stars.

Sanchez studies galaxies by using cosmological hydrodynamic simulations.

“I put a bunch of particles in a box, turn on gravity, and let time happen,” she laughed. After running a simulation she looks for a galaxy similar to the Milky Way, and examines interactions between the galaxy’s supermassive black hole and the circumgalactic medium.

“The supermassive black hole is actually really vital to the evolution of the CGM because it’s moving all of this metal that’s being created in the hearts of stars in the disk of the galaxy and it’s propagating them out into the CGM,” Sanchez explained. Without a supermassive black hole, the circumgalactic medium would not look like what astronomers have observed.

Pass the syrup.