After the Big Bang, the universe was a thin soup

Lex Fridman asks MIT astrophysicist Anna Frebel to go back before the Milky Way existed. Scientists believe there was a Big Bang, she says, but what matters for her work is which elements were left behind. The infant universe was hydrogen, helium, and tiny sprinkles of lithium. That was pretty much it, and it turns out to be a terrible recipe for structure: the gas was too hot to collapse, clump, or ignite fusion.

The first stars formed before any galaxies, from that pristine mix, and they were giants — about a hundred times the mass of the Sun. Big stars explode after only a few million years, a blink on cosmic timescales. In their cores they fused hydrogen and helium all the way up to iron, then dumped the products into space in supernova explosions. After that first blast, Frebel says, the universe was no longer chemically pristine, and that change set the stage for everything else, including the conversation they are having.

It took a few hundred million years for those first stars to emerge, then a few million more for them to go kaboom. Frebel likes to picture the universe as a nice soup. The first supernovae provided the salt: just a little sprinkle of heavier elements, and the flavor changed completely. Carbon, oxygen, magnesium, and iron appeared only in tiny amounts, but that was enough to alter the physics of the surrounding gas.

Why did that salt matter so much? Carbon and oxygen atoms are more complicated than a lone proton, and they can radiate photons out of a cloud, cooling it. Colder gas fragments into smaller clumps, and smaller clumps make smaller stars. A star like the Sun lives about ten billion years; anything less massive lives even longer, fifteen or twenty billion. That is the foothold of Frebel's career: a low-mass star that formed soon after the Big Bang can still be observed today.

Nuclear burning happens in the core. The outer layers barely talk to it, so the atmosphere keeps the recipe of the cloud the star was born from. Frebel does not have to look to the edge of the universe to study the beginning. She chemically analyzes the oldest nearby stars and unpacks what happened in the first billion years. When she was in high school, textbooks said the universe was 20 billion years old; the figure has since settled near 13.8 billion, she thinks, because cosmic microwave background maps and other independent data have mostly converged.

The earliest little stars, she guesses, formed maybe half a billion years after the Big Bang — a few hundred million years for the first giants, then time for the next generation. That is also when the first protogalaxies assembled, the stellar systems from which the Milky Way eventually grew. Galaxies grow hierarchically: the bigger ones eat their smaller neighbors. Those swallowed early stars now sit in the outer parts of our galaxy, which is why Frebel points her telescope at the outskirts.

A galaxy, Frebel says, is a huge assembly of stars. The Milky Way holds something like 200 to 400 billion of them, most of them in a disc. We live in a spiral disc galaxy. From the Northern Hemisphere the Milky Way band is the next outer spiral arm, with dark space behind it, so it looks relatively thin. Travel to South America and you see the next inner arm, backlit by the galactic center — a puffy region of gas and star formation that she calls the galactic party, which is why the southern Milky Way looks so much brighter.

If you ever see that sky, she encourages you almost to close your eyes and imagine sitting in the disc, in the pancake, looking right into it. You can feel that there is a top and a bottom, that we are really part of the galaxy rather than lost in empty space. Fridman asks whether that makes her feel small. She says she does not, not necessarily. She feels awe, and she feels she belongs. Many people picture a lonely planet and then nothing, which she finds almost sad. She prefers the image of a galactic merry-go-round: the Sun does circle the center, even if we cannot feel the ride.

Fridman wonders about the variety of worlds around other stars. Frebel notes that roughly every other star has at least one planet, which, given hundreds of billions of stars, is a huge census. Star and planet formation still resist detailed recipes: you need a cooling cloud, something clumps, and somehow a star appears, with or without planets. Local conditions vary the way people all look like people and yet look different. Planets themselves barely shine, so actually photographing one remains minuscule. There is a lot of possibility out there, she says; we have to be a little more patient.

Pretty much every large galaxy has a supermassive black hole in its center, and astronomers still do not know where those objects come from. In the early universe, Frebel says, dark matter structured up first and held gas and stars in shallow gravitational potential wells. Those first bound systems grew by absorbing even smaller neighbors. Somewhere in that process there must have been a seed for a supermassive black hole. She is not sure it is clear what was there first, the black hole or the galaxy.

A partial answer already exists close to home. The Milky Way is still surrounded by many dozens of dwarf galaxies, and Frebel has studied a bunch of them. To the extent astronomers can tell, those dwarfs do not contain black holes. Gravitationally bound structures — call them protogalaxies, dwarfs, or first galaxies — can exist without a central monster. Something was different about the bigger proto-Milky Way that let it attract and integrate the small ones. James Webb, the infrared space telescope, has been online about a year, and colleagues are staring at faint blobs whose light has traveled some 13 billion years, looking for the onset of those early black holes. There are already surprising results nobody really understands.

Frebel would like her JWST colleagues to push toward still older, more primitive structure. The technical ceiling arrives fast. Make the system smaller and it becomes dimmer; put it farther away and it is dimmer still. "We're reaching the end of the line from a technical perspective pretty quickly," she says. In that early phase of the universe, smaller does correlate with older. Past that phase, it does not.

Fridman asks how much crosstalk there is with the theoretical physics of black-hole paradoxes. At first glance, Frebel says, not much. She mostly observes stars. The physics is chemical and not extreme, even if fusion needs quantum tunneling. She likes talking with theory colleagues, and she has written papers with people who run galaxy simulations — not as far removed as the pen-and-paper black-hole crowd. Even then, they had the same interests and almost spoke two different languages.

It took real time to reach common ground. If scientists already have trouble talking to each other, she says, imagine talking to non-scientists. Everyone is interested in the same human mysteries, but from different angles. When you take the time, the process is interesting, and those mixed-language papers were some of her most successful, and the ones she enjoyed most. Fridman compares it to artificial intelligence and neuroscience: one community tries to engineer intelligence, the other to understand it in brains, from different data and backgrounds.

Simulators can set a pile of parameters and play God, spinning up universes to see which ones match the sky — Sims at cosmic scale. Observers get the nights the telescope assigns and collect real photons. Frebel calls the whole enterprise a big sandbox. Everyone has a corner, but they are in the same box. That box also contains super-expensive telescopes that the children fight over, because there is only one experiment, the universe, and what we see is what we get.

The less massive the star, the longer it lives. For dinner-party use, Frebel offers the Sun's ten-billion-year lifetime; at 0.6 or 0.8 solar masses you get 15 to 20 billion years. If such a star formed soon after the Big Bang, it is still observable. Fridman had spoken of old light. The light from these stars is only a few thousand years old, which compared with their ages is nothing. To her, that light is young. The stars sit in the outskirts of the Milky Way. They probably did not form here. Hierarchical assembly means they were eaten: born in a little neighbor, absorbed, and left on the fringe.

These dwarfs of stars are efficient. They have been burning hydrogen to helium in their cores for 12 or 13 billion years and will keep doing so, as the Sun will. Most of the star's gas never talks to the core, so the outer composition is exactly that of the birth cloud. Frebel is a stellar archaeologist: she does not dig in dirt for remnants of civilization. She digs for old stars because they have preserved the first billion years in their atmospheres. That is the best look at early chemistry anyone could wish for.

There is a caveat. They cannot accurately date these stars. The trick is chemical evolution. After the pristine Big Bang mix, the first stars sprinkled elements up to iron. The next generation included both exploding giants and long-lived dwarfs. Other processes — neutron-star mergers and exotic supernovae — made elements heavier than iron, all the way to uranium, from very early on. Frebel jokes that pretty much all the elements were done from day three. Iron is the baseline every supernova provides; gold, silver, and platinum are the spice, and she likes them for both scientific and personal reasons.

It took about a thousand generations, give or take, until the Sun was made. The Sun formed from gas enriched by roughly a thousand supernova explosions, which is why it has the composition it has, planets included. Each generation does not invent new species of atom up to iron. It just adds another sprinkle of the same set. Heavy elements beyond iron are not fusion products. They form through neutron-capture processes that need seed nuclei such as iron or carbon. Because those processes cannot run everywhere, a star loaded with heavy elements is telling you that something really interesting happened locally.

The old stars we observe formed from whatever had already polluted their clouds. Some, Frebel thinks, are second-generation: gas enriched by just one first star. Others carry a more complicated heavy-element signature — maybe a first star, a more normal supernova, and a neutron-star merger. The job is to unravel which processes, which events, and how many, produced the pattern we measure 13 billion years later. Amounts are so tiny that one ordinary supernova is already almost too much. That is how they sort second generation from third.

The thing you can never do in the universe, she stresses, is take things away. You can only add. There is no cosmic vacuum cleaner. Black holes are the closest thing, but they would have taken the whole star, not metals out of the gas. Maybe ten stars now contain so little of the heavy elements that they must be second generation; how else would you have made them? And a clarification she loves to repeat: the elements in these stars were not made by the stars themselves. They are a reflection of the cloud. "At the end of the day, we don't really care for the stars that we're observing, we care for the story that they're telling us about the early universe."

These days Frebel pairs elemental signatures with kinematics — how a star moves through the galaxy — for clues to where it came from. All these old stars are in the galaxy, but they are not of the galaxy. A few days before the conversation she submitted a paper with three women undergraduates. They found stars with very low strontium and barium, the heavy elements that flag special events. Because you can only add, an extremely low abundance means at most one such event, either very early or in a very special environment.

Then they looked at the motions. The stars are all going the wrong way. The protogalaxy, she says, was like a beehive that did not know what it wanted to become. It was absorbing little galaxies to grow fast; some came in with the flow and some the wrong way. That can only happen early, before there was a left and a right and an up and a down. Thirteen billion years later those stars are still doing it, at hundreds of kilometers per second. Retrograde motion is a clear sign of accretion. You cannot throw a galaxy into the Milky Way the wrong way today; it will eventually turn around.

The collaboration grew from a new MIT class. Frebel wanted to stop requiring a cold email to a professor. She mapped an entire research project into ten worksheets so students could sign up in teams of two, each getting an old star that had never been analyzed, with no solution at the back of the book. Anyone can crunch numbers, she says; the scientist's job is to find the story from three puzzle pieces. Nine people signed up, eight of them women. They were so into it that she invited extra work toward a paper. With colleagues she talks about chemically most pristine stars, not "oldest," because abundance is what they measure. These low-strontium, wrong-way stars, she argues, must be among the oldest that came in before the Milky Way was the Milky Way.

Astronomers long ago decided to call hydrogen X, helium Y, and everything else metals, Z. Frebel likes to tell public audiences that neon is a wonderful metal, to the horror of any chemists present. Chemical evolution is then a monotonously increasing function of those "metals." Searching for the most metal-poor stars is walking that function backward toward the first stars. After the pristine post-Big Bang mix, the mess started, and the mess became today's variety. The zoo of metal-poor stars is already stark, and some crazy patterns still will not fit.

Two of her "baby stars" are HE 1327-2326 and HE 1523-0901, names built from old coordinates: the minus puts them in the Southern Hemisphere. Colleagues at conferences have called them Anna star or Frebel star to avoid learning the phone number. HE 1327 was the second second-generation star found. The first can be a quirk; the second showed the search techniques work. Both had huge amounts of carbon relative to iron. Iron is the usual reference for metallicity. Their iron was far below what a normal supernova should have dumped, and you cannot take iron away. Population III, the first stars, must have exploded differently than theorists had assumed.

Japanese colleagues cooked up a fainter supernova without enough oomph to eject much iron — and then had to explain a thousand times more carbon. The established picture is a fallback mechanism: during the explosion a massive black hole forms, inner iron falls back, and carbon farther out escapes. A massive star before death is layered, hydrogen and helium outside, an iron core in the center. Fusion of iron yields no energy, so the star implodes, bounces, and disrupts. Carbon from those first stars, Frebel argues, probably cooled the next clouds enough to make the first low-mass stars. If carbon's properties had been different, there might be no long-lived planet hosts, and we are carbon-based besides. "We're made of star stuff," she says. Carl Sagan got it right. Biological evolution on Earth was facilitated by the chemical evolution of the universe.

HE 1523-0901 is a nearby red giant, not as iron-poor, maybe third to fifth generation. What mattered is a huge overload of very heavy elements, including measurable thorium and uranium. Thorium's half-life is about 14 billion years, uranium's about 4.7, timescales useful for the early universe. Those nuclei come from the rapid neutron-capture process, the r-process: a seed such as iron is bombarded with neutrons faster than it can beta-decay, fattened past uranium, then left to decay to the stable isotopes we know. The whole thing, Frebel says, is done within two seconds. In talks she shows a simulation that is slower than real time — the opposite of the usual 13-billion-year galaxy in 30 seconds.

Where do you find that many neutrons? Neutron stars, leftovers of supernovae, often in pairs. Their orbit decays until they merge in an explosion with roughly supernova energy but different physics, sending ripples through spacetime. In 2017, LIGO and Virgo detected such a merger; telescopes found an electromagnetic counterpart that faded over two weeks, the light curve of heavy nuclei decaying to stability. That confirmed neutron-star mergers as one r-process site. A year earlier, in 2016, Frebel and former graduate student Alex Ji found the ultra-faint dwarf Reticulum II, still orbiting the Milky Way, full of ancient iron-deficient stars with the same strong r-process signature as HE 1523. They had gone looking for the opposite: the low strontium and barium they thought was typical of early star formation.

The stars are still in their native galaxy, so the environment is known. Frebel inferred an early neutron-star merger that polluted the gas. Ultra-faint dwarfs may hold only a few thousand stars. Reionization, plus a few supernovae in a shallow well, blew or evaporated their gas, so they have not formed stars since. For a stellar archaeologist that is a gift: a whole population waiting in place. Milky Way halo stars are brighter, so the spectra are better — HE 1523's uranium detection needed that. Dwarf stars give site constraints instead. A theorist might say stack a hundred supernovae to build the r-process yield. Frebel answers that this dwarf is still intact; five or ten supernovae would have blown it apart. That is why the system being alive matters.

Frebel mostly uses the Magellan telescopes in Chile, 6.5-meter mirrors with an unusually efficient spectrograph that helps make up for not being eight meters. She is a spectroscopist; she does not take pictures. Before the pandemic, Magellan was among the last places where astronomers still traveled to take their own data. Most observatories now have staff do it, which is convenient and, she thinks, strips out a central part of the work. There is no apparatus in the basement. You go to a remote mountaintop, wait for sunset, and collect photons. During 45- or 55-minute exposures of faint dwarf-galaxy stars she has run outside, lain on the ground like a snow angel under the southern Milky Way, and felt she belongs rather than feeling small. In that dark site you can walk in starlight and almost see a shadow.

A spectrum is a rainbow made with a prism instead of a raindrop: a wriggly line with dips, black lines eaten out of the colors. You only ever see the stellar surface. Atoms there absorb specific photons from the hot core, and the strength of each dip says which element and how much. The solar spectrum is a tangle of overlapping lines. Old metal-poor stars are almost a straight continuum with occasional barcodes — iron here, carbon, magnesium, a little strontium. Codes that mimic stellar atmospheres turn those equivalent widths into abundances. At the telescope she takes a 10- or 15-minute quick look, processes a summary plot of key lines, and reads the tea leaves: thumbs up or thumbs down. She now does this from her living room over Zoom. Searching is a numbers game, starting from millions of stars. Her supervisor's saying: one star is a discovery, two is a sample, three is a population.

The most iron-poor stars have lines almost in the noise. You are looking for the last detectable bit of nothingness, but if there is truly nothing you get only an upper limit. The record holder still has only an upper limit. False positives include white dwarfs, which have lost their atmospheres and can impersonate metal-poor stars if you miss their slightly wider hydrogen lines. Heartbreak can be three cloudy nights after you won the time. The Reticulum II run was worse: a student called at 2 a.m. convinced he had observed the wrong star because of a europium line that looked metal-rich. It was the first r-process galaxy, unpredicted. Weather was collapsing, the system was setting, and they would not get another window for nine months. By a hair they got the nine brightest stars in atrocious data. The enhancement was so strong it stuck out of the noise. If it had been milder, they could have said nothing. Frebel did not sleep much that night.

Frebel is excited about JWST and is even on a proposal or two, but she is not planning to use it as her main tool. She and the JWST teams want the first billion years from opposite sides of the fence: she has surviving stars and the stories they imply; they catch the earliest photons and the energetics. Stellar archaeology has filled in a canvas that was nearly blank twenty years ago, when the first second-generation star was a serendipitous find. Individual 10-sigma discoveries are getting rarer; if they were easy, they would already have been made. Finding another second-generation star will not move the needle; finding 50 or 100 might. Large spectroscopic surveys are coming, and the hand-grown, one-star-at-a-time field she loves is about to change. She still wants the last pieces of a thousand-piece puzzle, not just the broad picture. Close one door, she says, and ten new ones open.

Stars cannot probe before they formed, so they will not explain the Big Bang itself. Thorium and uranium in HE 1523 gave a best-effort age of 13.2 billion years, an average of elemental ratios that individually said 15 or 10. Fifteen is a correct calculation inside the model and nonsense if the universe is 13.8 billion years old. The physicist's job, she tells students, is to take the math and ask whether it makes sense. At the Big Bang we no longer know what is supposed to make sense, so math may go further than physics, judgment-free, even into extra dimensions. Her light has traveled only a few thousand years; JWST's has traveled ~13 billion. Those old stars are alive and kicking, but they are poor vacation spots: planets need heavy elements, and the universe took billions of years to make enough of them.

Her book highlights women who were not immediately credited. About a century ago the Harvard Computers — women hired for low wages and patient work — combed the big data of the day and made discoveries anyway. Annie Jump Cannon classified thousands of spectra into the temperature sequence still used. In 1925 Cecilia Payne-Gaposchkin showed that the Sun is mostly hydrogen and helium, against the assumption that celestial objects match the Earth; the thesis was later called among the most important of humankind. Marie Curie found radium and polonium. Lise Meitner's nuclear fission, calculated in 1938–39, is the basis for understanding r-process fission cycling; she was nominated for the Nobel more than forty times and never received it. Frebel, who is from Germany, later wrote a 25-minute one-woman play in which she portrays Meitner's discovery, then turns into herself for a 30-minute public talk on the r-process — a hundred-year torch pass the papers never describe, because journals serve results on a golden platter and skip the moment you found the car keys.

Asked for the most beautiful idea in stellar astronomy, she describes sitting in a talk in Australia and realizing that old stars combined the stars, the elements, and the nuclear physics she had not known how to join. Twenty years on, she feels she won the lottery. Her advice to young people is to put eggs in a basket: abundance of options can make people too tired to choose, afraid of missing the other 99 things, when life is always missing 99 things. Dig in on one thing for a time, with heart and hands. Opportunity is nothing if you do not do anything. On meaning, she notes that biological evolution required chemical evolution, with carbon in a starring role. We may be a consequence of the laws of physics and the local conditions around the Sun. She likes believing that, because then it is reproducible elsewhere; pure chance is not truly satisfying to a scientist. Fridman calls it turtles all the way down. Frebel agrees. She has studied some of the most ancient turtles at the bottom, and they live for quite a while.