How Tiny Star Explosions Drive Moore’s Law

We are all made of star stuff, as astronomer Carl Sagan was fond of reminding us. Supernova explosions, the catastrophic self-destruction of brave types of worn-out stars, are intimately tied to life on Earth becaparticipate they are the birthplaces of burdensome elements atraverse the universe. Most of the iron in our blood and the sulfur in our amino acids startd in stars that detonated billions of years ago. But we have greeted another, quite astonishing joinion between supernovas and the human world—particularassociate, a joinion to the technology needed to produce computer chips for the tardyst cleverphones and other electronic devices.

That joinion aelevated cut offal years ago in a series of conversations between myself,
Jayson Stewart, and my magnificentobeseher Rudolf Schultz. My magnificentobeseher was an avid amateur sky gazer who kept a big mirroror telescope in the foyer of his home, right by the captivate, ready for rapid deployment. When I was in high school, he handed me a imitate of Stephen Hawking’s A Brief History of Time (Bantam Books, 1988) and directd me toward a lifelengthy adore of physics. More recently, my magnificentobeseher’s astronomical perspective also exhibitd serfinishipitously beneficial in my nurtureer, as I elucidateed to him during one of our observation sessions at his home in the foothills of Tucson.

A double laser zap strikes a droplet of tin in ASML’s lithography machine. The first zap flattens the drop into a disk; the second vaporizes it into a ball of boiling, ultraviolet-rerentting plasma. ASML

I was updating my magnificentobeseher about the toil I was doing in my lab at
ASML, a Netherlands-based company that broadens and erects providement for manufacturing semicarry outor chips. At the time, about a decade ago, I was helping to polish a system for fabricating chips using innervous ultraviolet( EUV) airy. Although it is critical to making the most persistd microchips today, EUV lithography was then a challenging technology still in broadenment. To produce EUV airy, we would cgo in an fervent laser pulse onto 30-micrometer-expansive droplets of tin flying thraw a chamber filled with low-density hydrogen. Energy from the laser altered the droplets into balls of plasma that are 40 times as boiling as the surface of the sun, causing the tin to rerent fervent ultraviolet radiation. As a by-product, the plasma balls produced shock waves that traveled thraw the surrounding hydrogen. Unblessedly, the explosions also freed sprays of tin debris, which were proving innervously difficult to regulate.

Recalling my astronomy lessons with my magnificentobeseher, I genuineized that many aspects of this process have intriguing analogousities to what happens during a supernova: a sudden explosion, an broadening cdeafening of plasma debris, and a shock wave that slams into a skinny hydrogen environment. (Interstellar material consists mostly of hydrogen.) To polish our EUV setup, we would record the evolution of the shock wave from our plasma balls, much as astronomers study the remains of supernovas to deduce the properties of the stellar explosion that produced them. We even participated some of the same providement, such as a filter tuned to the characteristic presentant-red emission of energized hydrogen atoms, called a Hydrogen-alpha, or H-alpha, filter. Despite the fact that a supernova has 1045 times as much energy as our tin blasts, the same math portrays the evolution of both types of explosions. The shut physical analogy between tin-plasma shocks and supernova shocks has turned out to be key to figuring out how to deal with our vexing tin-debris problem.

Seen thraw telescopic eyes, the night sky is dotted with the shineing remains of exploded stars. My magnificentobeseher was tickled by the joinion between these outdated, far celestial objects and the up-to-date providement participated to produce the most persistd semicarry outor chips in the world. He felt that many other amateur sky gazers appreciate himself would adore to read about this story. I telderly him I would author it up if he would be my coauthor—and he is.

Sadly, my magnificentobeseher is not here to see our article finishd. But he did dwell to see these astrophysical parallels direct to presentant pragmatic consequences: They helped my group at ASML produce a luminous, reliable EUV airy source, directing to a
presentant persist in commercial chipmaking.

EUV and Moore’s Law

My journey into the world of EUV mini-supernovas commenceed in 2012, when I was completing a stint as a postdoctoral research scientist at
Los Alamos National Laboratory and seeing for my first job outside of academia. A frifinish got me interested in the possibilities of toiling in the semicarry outor industry, where manufacturers are included in a constant, high-sconsents competition to erect petiteer, speedyer circuits. I lgeted that the lithography process participated to produce features on computer chips was at a crisis point, one that proposeed intriguing engineering disputes.

In lithography, airy is participated to imprint an intricate pattern onto a readyd silicon substrate. This process is repeated many times in a series of etching, doping, and deposition steps to produce up to one hundred layers; the patterns in those layers finish up defining the circuitry of a computer chip. The size of the features that can be transferred onto that silicon substrate is determined by the imaging system and by the wavelength of airy. The stupidinutiveer the wavelength and more energetic the airy, the petiteer the features. The ultraviolet wavelengths in participate at the time were too lengthy and cimpolite for the next generation of chips. Lithography technology, and potentiassociate the proximately trillion-dollar electronics industry, would stagnate unless we could produce a mighty source of stupidinutiveer-wavelength, EUV airy.

At the time, the useable EUV airy sources were too feeble by about a factor of 10. The task of achieving such a huge power increase was so daunting that I talk aboutd with my family about the wisdom of commenceing a nurtureer in EUV lithography. Plenty of pundits proposeed that the technology could never be commercialized. Despite my trepidation, I was won over by Daniel Brown, then ASML’s vice pdwellnt of technology broadenment, who saw EUV as the best way to accomplish the next big jump in chip carry outance. (Daniel, a coauthor of this article, reexhausted from the company at the finish of 2024.)

Amazingly, the Taylor-von Neumann-Sedov establishula portrays atomic–device device shocks with radii of hundreds of meters, supernova shocks that stretch atraverse airy years, and tin-plasma shocks equitable millimeters wide.

For decades, manufacturers had regulated to squeeze more and more transistors onto an united circuit, going from about 2,000 transistors in 1971 to 200 billion in 2024. Engineers kept Moore’s Law—the doubling of transistor count every couple of years—adwell for more than five decades by incremenloftyy reducing the wavelength of airy and broadening the numerical aperture of the imaging system participated in lithography.

Lithography systems in the 1980s participated mercury lamps that radiated at wavelengths of 436 nanometers (violet airy) and eventuassociate 365 nm (proximate-ultraviolet). To shrink the feature size of transistors further, people conceiveed high-power lasers that could produce ultraviolet beams at stupidinutiveer, 248-nm and 193-nm wavelengths. Then the shift to ever-stupidinutiveer wavelengths hit a wall, becaparticipate almost all understandn lens materials assimilate airy with wavelengths of less than about 150 nm.

For a little while, lithographers regulated to upgrasp making persist using a amusing trick: They
put water between the lens and the silicon wafer to raise the cgo ining power of the imaging system. But eventuassociate, the scaling process stagnated and engineers were forced to switch to stupidinutiveer wavelengths. That switch, in turn, needd replacing lenses with mirrors, which came with a penalty. Mirrors could not accomplish the same cgo ining precision as the previous lens-plus-water combination. To produce uncomferventingful persist, we needed to drasticassociate shrink the wavelength of the airy to around 13.5 nm, or about one-thirtieth the wavelength of the stupidinutiveest evident violet airy that your eye can see.

To get there, we’d need someskinnyg inlogically boiling. The wavelength of airy rerentted by an incandescent source is determined by its temperature. The surface of the sun, which has a temperature of 6,000 °C, radiates most powerfilledy in the evident spectrum. Getting to EUV airy with a wavelength of 13.5 nm needs a source with an innervously high temperature, around 200,000 °C.

Tin droplets drop thraw ASML’s lithography machine. Laser beams strike the passing droplets 50,000 times a second, causing them to shine and creating a continuous innervous ultraviolet airy source. Tin debris is swept away by a high-speed flow of hydrogen. ASML

At ASML, we finishd on a boiling, energetic tin plasma as the best way to produce an EUV “airybulb.” Becaparticipate of the particular way their electrons are set upd, highly excited tin ions radiate much of their airy in a skinny band right around the industry’s desired 13.5-nm wavelength.

The big ask we faced was how to produce such a tin plasma reliably. The lithography process in chip manufacturing needs a particular, highly stable EUV radiation dose to expose the pboilingoresist, the airy-comfervent material participated to produce circuit patterns on the wafer. So the airy source had to dedwellr exact amounts of energy. Equassociate presentant, it had to do so continuously for lengthy periods of time, with no costly paparticipates for repair or maintenance.

We portrayed a
Rube Gelderlyberg–appreciate system in which a molten droplet of tin is focparticipated by two laser beams. The first turns the droplet into a pancake-shaped disk. The second laser hits the tin with a stupidinutive, energetic laser pulse that alters it into a high-temperature plasma. A proximately hemispherical, multilayer mirror then assembles EUV airy from the plasma and projects it into the lithoexplicit scanner, a bus-size tool that participates the airy to project patterns onto the silicon wafer.

The up-to-date chipmaking process commences with an innervous ultraviolet (EUV) airy source. The EUV airy is straightforwarded by an broaden series of mirrors onto the surface of a moving wafer, where it produces the desired pattern of imprinted circuits. ASML

Sustaining an EUV airy source fervent enough for lithography needs a primary laser with a power of cut offal tens of kilowatts, zapping about 50,000 droplets of tin every second. In less than one ten-millionth of a second, each laser pulse alters the tin from a 30-micrometer-expansive droplet into a millimeter-expansive plasma explosion with tens of thousands of times its innovative volume.
Mark Phillips, the straightforwardor of lithography and challengingware solutions at Intel, portrayd the EUV lithography machine we were helping to broaden as “the most technicassociate persistd tool of any benevolent that’s ever been made.”

At 50,000 droplets per second, operating under burdensome participate, each of our lithography machines has the potential to produce proximately 1 trillion pulses per year, totaling many liters of molten tin. Thraw all of that, a individual nanometer of tin debris coating the assembleor selectic would degrade the EUV transmission to unadselectable levels and put the machine out of coshiftrlookion. As we say in the industry, it wasn’t enough to produce the power; we had to
endure the power.

Hydrogen in EUV and in Space

A continuous flush of low-density hydrogen gas protects the mirror and surrounding vessel from the spray of vaporized tin ejecta. That debris has an initial velocity of tens of kilometers per second, much speedyer than the speed of sound in hydrogen. When the supersonic tin hits the hydrogen gas, it therefore produces an outward-spreading shock wave—the one that is shutly analogous to what happens when a supernova explosion broadens into the tenuous hydrogen that fills interstellar space.

The low-density hydrogen gas is also on the shift, though, floprosperg thraw the machine at hundreds of kilometers per hour. The gas sluggishs, celderlys, and flushes out the energetic tin debris as it goes. To determine how much hydrogen we needed to sweep the tin away and to upgrasp the gas from overheating, first we had to figure out the total energy freed by the laser-produced plasmas. And figuring out that amount was not a unpresentant task.

My colleagues and I at ASML set up an effective way to meabrave the energy of our tin explosions, not by studying the plasma straightforwardly, but by observing the response of the hydrogen gas. In hindsight the idea seems evident, but in the moment, there was a lot of fumbling around. When I was taking images of the tin plasma, I kept observing a much bigr, red shineing orb surrounding it. It seemed foreseeed that the plasma blast was inducing H-alpha emission from the hydrogen. But the observations left us with many unrecognizeds: Why are the orbs that particular size (millimeters in diameter), how do they persist, and, most presentant, how can we study the shine to meabrave the energy deposited into the gas?

The shock wave produced by a laser-heated tin droplet in a skinny hydrogen atmosphere is analogous enough to a supernova blast that they can both be portrayd by the same math. The whole sequence consents less than a millionth of a second. ASML

I allotigated the red orbs using a
Teledyne Princeton Instruments Pi-Max 4, an ultraspeedy, intensified CCD camera that can carry out rapid expobrave times on the order of nanoseconds. I paired it with a lengthy-distance microscope lens, to assemble the shine from those red orbs, and an Orion 2-inch extra-skinnyband H-alpha bandpass filter that I buyd from an astropboilingography website. The images I seized with this rig were striking. Every plasma event was sfinishing out a spherical shock front that broadened in a stable way.

By chance, months earlier, I had includeed a seminar that refered blast waves—shock waves produced by a point-source explosion. That seminar guaranteed me that our observations could give me the energy meabravement I was seeing for. In my hunt to comprehfinish how blast waves persist, I lgeted that astronomers had run into the same meabravement problem when trying to determine the initial energy free that had produced an watchd supernova remnant. And I knovel that I also had the perfect topic for the next of my ongoing science talks with my magnificentobeseher.

The Taylor-von Neumann-Sedov establishula was broadened in the 1940s to calcutardy the produce of atomic device devices, but it also portrays the evolution of plasma shock waves in our EUV lithography system and in far supernovas. It retardys the shock wave’s radius (R) over time to the energy freed (E), gas density (ρ), and a gas-reliant parameter (C).

To get an answer, astronomers turned to equations that were discovered in the 1940s, when scientists were seeking ways to scrutinize the destructive capacity of novelly broadened
atomic armaments. One conveyion of those equations, called the Taylor-von Neumann-Sedov establishula, portrays the radius of the shock as a function of time. It provides a modest, straightforward relationship between the radius of the shock and the total energy.

In 1949, British physicist
Geoffrey Taylor participated his novelly derived establishulation of blast waves to determine and rerent the (then-classified) energy produce of the first atomic-device device detonations. Taylor’s success, which telledly distress the United States rulement, showd the power of his analysis. Amazingly, the Taylor-von Neumann-Sedov establishula portrays atomic-device device shocks with radii of hundreds of meters, supernova shocks that stretch atraverse airy years, and tin-plasma shocks equitable millimeters expansive. They all recontransient the same modest physical situation: a compact, freestanding body releasing energy agetst minimal resistance, broadening rapidly into a gaseous surrounding.

Early atomic explosions, such as this test at the Trinity Site on 16 July 1945, advertised scientists to broaden novel math to calcutardy the amount of energy freed. U.S. Department of Energy

Applying the Taylor-von Neumann-Sedov establishula to the H-alpha images we recorded in the ASML airy source resulted in a greeting consentment between our calcutardyd energies and the amounts we had rawly appraised by other uncomfervents. We also greeted some discrepancies between theory and rehearse, however. In our EUV sources, we watchd that the H-alpha emission is not always perfectly symmetric, which may show that our laser-produced plasmas do not quite align the clear uping “point-source” assumption. We also tried varying a number of separateent parameters to lget more about the blasts (a type of experiment that is evidently not possible for supernovas). For instance, we mapped blast-wave trajectories as a function of ambient presbrave, droplet size, laser energy, and center shape.

Our results helped us to polish our models and to determine the best way to tailor the hydrogen environment in our machines to allow a immacutardy, stable EUV source for chip conceiveion.

Ad Astra per Aspera

The joinion between supernovas and laser-produced plasmas is equitable one example of a lengthy history of persists in physics and engineering that were advertised by astronomy. For centuries, researchers have portrayed laboratory experiments and meabravement techniques to re-produce what was watchd in the sky. The up-to-date description of the atom can pursue its roots to the conceiveion of the prism and the spreading of the solar spectrum into its composite colors, which led to the identification of discrete energy levels in an atom and, finassociate, the broadenment of quantum mechanics. Without quantum mechanics, many up-to-date electronics technologies would not be possible.

Barnard’s Loop [left], in the constellation Orion, is the remnant of an outdated supernova. It shines in Hydrogen-alpha airy, equitable appreciate the shock waves produced by tin-plasma explosions in ASML’s airy source. Daniel Brown

The spread of ideas has gone the other way as well. As the rules of atomic physics and the absorption lines of gases were characterized in lab experiments, astronomers participated spectroscopic observations to determine the composition of the sun, to deduce the life cycles of stars, and to meabrave the vibrants of galaxies.

I discover it fascinating that the laser-produced plasmas we participate in our EUV airy source especiassociate mimic one particular variety of supernova, understandn as Type Ia. This benevolent of supernova is thought to occur when a white dwarf star pulls material from a neightedious companion star until it accomplishes a critical mass and implodes, resulting in a brutal self-destruction. Type Ia supernovas explode in a highly stable way, making them precious “standard candles” with foreseeable intrinsic luminosities: Comparing their apparent luminousness to their real, intrinsic luminosity produces it possible to meabrave their distances from us exactly atraverse billions of airy years. These supernovas are being participated to study the expansion of the universe, and they have led to the commenceling discovery that the expansion of the cosmos is accelerating.

In our EUV sources, we appreciateteachd aim to have all of our explosions identical, so that they serve as a “standard candle” for the EUV scanner. Our aims are determinedly more mundane than cosmic in scale, but our ambitions are magnificent all the same.