Useful News for You

In early 2016, a team led by Capasso, with key contributions by Khorasaninejad, graduate student Rob Devlin, and postdoc Wei Ting Chen, showed that it indeed could be done and done well enough that commercial devices were possible. Capasso filed a report of invention with the Harvard Office of Technology Development and, soon after, the discovery made Science’s cover. Gordon, OTD’s director of business development for physical sciences, stepped in to manage the avalanche of interest.

“I’ve been doing this for far longer than I like to admit but that paper, the invention, and the patent we filed generated far more commercial interest — from companies, entrepreneurs, investors — than any other hard-tech invention I can remember,” said Gordon. “It was exciting and a bit shocking. We met and talked with a lot of people about this work.”

Those people understood then what Capasso had seen more than a decade earlier. Lenses are essential components in a host of devices, focusing and detecting light — both visible and invisible — for applications well beyond imaging, including facial recognition in smartphones and laptops, proximity and gesture detection to enhance responsive functions in automated devices, depth-sensing cameras, environmental awareness in drones and robots, and collision avoidance in self-driving cars.

In many of those devices, space is tight. The stacked elements of plastic or glass in traditional lenses have resisted the true miniaturization that most other components have undergone. They remain among the bulkier components, and a bottleneck in device design.

“I hold up my cellphone and pull out a credit card,” Gordon said, describing how he introduces the technology to potential investors. “There are only two reasons the phone is not as slim as the credit card. One is the camera and the other is the battery. The metalens will help enable the phone to be as slim as a credit card.”

While traditional lenses use curved glass or plastic to bend light and focus an image, a metalens uses a series of tiny pillars on a millimeter-thin wafer. The pillars are smaller than the wavelength of light and transparent to the desired wavelength. The pillars’ shape, the distances between them, and their arrangement on the wafer are varied to bend light as desired.

Not long after that Science cover, OTD licensed the technology to a startup, Boston-based Metalenz, founded by Capasso, Devlin, and Bart Riley, a tech entrepreneur with whom that office had previously worked. Now Metalenz’s chief executive, Devlin made a key materials advance in the Capasso lab that greatly improved the lenses’ efficiency. Earlier this year, Metalenz logged its first major sale, with manufacturer STMicroelectronics. STMicro will use metalenses in the company’s “time of flight” modules, which provide 3D sensing in an array of devices and which have previously sold 1.7 billion units. Those units appear in everything from drones to robots to smartphones. Metalenz said in June that it expects its optical components to be in millions of consumer devices this year.

Khorasaninejad, who today is CEO and cofounder of San Francisco-based Leadoptik, called the deal “a very, very strong endorsement from industry,” while Capasso said that the metalens can be made in the same factories as computer chips is potentially “game-changing,” as it unifies two industries: semiconductor manufacturing and lens-making.

“The same planar technology, known as deep ultraviolet lithography, to mass-produce integrated circuits — chips — can be used by the same foundry to make flat optics such as metalenses,” Capasso said. “It means that the entire camera module of a cellphone or laptop will eventually be manufactured in one sweep, including the metalens and the sensor.”

‘Can you get rid of the lens?’

Capasso came to Harvard in 2003 after a career at Bell Labs, where, in 1994, he and colleagues invented and developed the quantum cascade laser, currently being commercialized in devices for chemical sensing and spectroscopy.

Capasso traced the development of the metalens to a conversation he had more than a decade ago with Jim Anderson, the Philip S. Weld Professor of Atmospheric Chemistry. The two had been discussing putting a quantum cascade laser on a drone that Anderson wanted to use to detect certain chemicals in the atmosphere, but there wasn’t enough room. That was in part because of the bulky optical elements needed for focusing. Anderson got to the heart of the problem.

“He said, ‘Can you get rid of the lens?’” Capasso recalled. “My first reaction was, ‘What the hell is he talking about?’ But then I said, ‘No, wait a moment.’”

Capasso started to brainstorm the idea with a couple of students in his group. Starting in 2007 or 2008, they began to focus on the scientific question of whether it was possible to bend light in an entirely flat device.

Early work used what they termed “plasmonic antennas” that eventually evolved into metasurfaces — millimeter-thick, two-dimensional surfaces studded with tiny nanostructures smaller than the wavelength of light. Those arrays, Capasso said, can alter the path of light flowing through it in a kind of “artificial refraction.”

That work progressed incrementally, producing several scientific papers that led to a 2011 breakthrough, published in Science and now cited more than 5,400 times. Capasso and members of his lab demonstrated they could tune the nanostructures and bend light to a roughly focused “hotspot.”

While the results were of scientific interest, the efficiency was so low that most of the light was lost, a result that skeptics said meant it would never become useful.

Bringing fresh eyes, new skills

In 2012, Devlin joined the lab. With no optics background, Devlin instead worked in an area Capasso thought would be key: materials science and nanofabrication. Devlin himself believed that the lab had mostly figured out the science, and that choosing the right materials and fabrication processes would be critical to improved results.

“The metasurface was a great proof of concept, but the devices themselves were really inefficient,” Devlin said. “There were problems in how it was fabricated, in materials, and design.”

Devlin set about considering materials and processes that not only worked in the lab, but that would also work, if a successful device needed to be scaled up. Ultimately, he settled on titanium dioxide, a compound widely used in paint pigment, sunscreen, food coloring, and as a reflective surface in dielectric mirrors. More importantly to Devlin, it had low light-absorption properties.

By late 2015, there were eight to 10 people in the lab working on different aspects of metasurfaces. As each turned their focus to lens performance, they brought perspectives and insights gleaned from their diverse prior efforts.

Devlin knew they had things right when the efficiency — the amount of the available light the device could focus — abruptly began to climb, rising from 10 percent to 85 percent in a few weeks. The rapid improvement and clarity of the resulting images stunned Devlin, Khorasaninejad, and Chen.

Those results spurred the 2016 Science paper, which not only generated a buzz in the lens industry, but also became a runner-up for Science’s breakthrough of the year. Despite the accolades and new belief in the promise of a metalens, it still focused just single wavelengths of light. And, while there were certainly uses for single-wavelength light — facial recognition, for example, is done by bouncing a single wavelength outside the visible spectrum off a person’s face and analyzing the light that returns — the next challenge was to produce the first “achromatic” metalens, one that focuses light across the visible spectrum.

“I locked myself in my office with Wei Ting Chen for the weekend, and I said, ‘Now we need to understand what we have done,’” Capasso said. “Our design ensured that all the colors, irrespective of where they take off, arrive at the same time in the same spot.”

It took two years, but in 2018, they became the first to report success, with high resolution.

In the meantime, though, Gordon was fielding calls from industry and advising Capasso and Devlin as to the next step of the metalenses’ commercial development. He counseled them that founding a startup around a new technology tends to be more successful than licensing it to a large corporation, where it can get lost. They listened and founded Metalenz to commercialize the invention and look for additional applications.

Devlin, meanwhile, had a decision to make. He had entered his graduate studies thinking he would pursue an academic track when he left Capasso’s lab. But he had the opportunity to be among the founders of Metalenz and shepherd the device’s development himself.

To Capasso, though, Devlin’s first priority was finishing the degree. He didn’t let up on the younger scientist, requiring he continue research and complete another scientific paper. Then Capasso ran interference with investors and companies wanting a piece of Devlin’s time.

“Federico made sure I was not abandoning the completion of my Ph.D.,” Devlin said. “He said, ‘No one is to contact Rob until he completes his dissertation.’”

Devlin defended his dissertation in 2017 and graduated in June 2018. When OTD and Metalenz announced the startup to the world in 2021, Devlin was its chief executive.

“The Ph.D. student is not always a CEO type, but some are, and Rob has shown he has the personality for it,” Gordon said.

The last several years have seen Devlin taking the company through several startup milestones, securing funds, developing relationships with manufacturers, and the recent announcement that the company’s metalenses will go into mass production.

Though Metalenz has licensed more than 20 Capasso lab patents from Harvard, Capasso and his students and postdocs are busily working on new discoveries. A current focus is on ultracompact, polarization-sensitive cameras — based on flat optics — that can detect the direction in which light vibrates after it’s transmitted or reflected, revealing otherwise invisible details of a scene. His group is involved with two collaborations with NASA on these cameras, related to Earth sciences and solar physics. He and his students are also toying with the idea of a “universal camera” that can see all properties of light at the same time, including ones that can’t be seen by existing cameras. Capasso described that challenge as “very ambitious.”

“We are here to learn, starting with me,” Capasso said. “I always tell students, ‘If you have something good, you have to give it away. Don’t keep it to yourself.’”

GAZETTE: There are a lot of incentives aimed at consumers in the legislation, with tax credits for energy efficiency and electric cars. Does it take largely a bottom-up approach or is it more balanced than that?

HOLDREN: I think it’s more balanced. It’s both bottom-up and top-down. I’ve always thought that comprehensive programs on almost anything significant — and difficult — need to be a combination of bottom-up and top-down. This looks to me like a pretty good balance. There are important things missing that were in the other measures the Biden administration was trying to get through, but there were also some not-so-good elements in there that have been dropped. I don’t argue that this is an ideal piece of legislation, but it is so much more than we thought we were going to get.

GAZETTE: Are there specifics in there that you like in particular? That may have multiplier effects?

HOLDREN: The energy and climate total, $385 billion, is terrific. The clean-manufacturing tax credits represent a very productive approach. The tax credits to consumers for buying electric vehicles and energy-efficiency improvements, too. We know that those work. The real name of the game in getting emissions under control is making the clean options more attractive economically than the dirty options. It’s that simple. So any measure that can contribute to creating or expanding the economic advantage of going with clean and efficient technologies is a smart thing to do, and there’s a lot of that in here.

There are also some specifics, like very substantial incentives to companies for reducing methane emissions, that are a big deal. Some of the measures focused on offsetting the economic costs are also very good. There is always an emphasis on how much it’s going to cost to have cleaner options faster, but there is almost never a discussion of how much money those measures will ultimately save the economy through reduced damages from climate change. I wish there was more attention paid to the economic benefits of taking these steps. Our late, great economics Professor Dale Jorgenson, who died recently, was a great exponent of the proposition that doing what we need to do to address climate change would ultimately be an economic benefit, not a cost.

GAZETTE: Looking at cost, will we eventually reach a tipping point after which renewables will be cheap enough that they’ll spread on their own, like natural gas did when it undercut coal? Is it possible that this legislation can get us to that tipping point?

HOLDREN: In some respects, we’ve gotten to that point already. Electricity generation from solar photovoltaics and from wind power in many places is now cheaper than electricity generation with coal. And in some places, it’s cheaper than electricity generation with natural gas. That’s a very important driver. One of the challenges is that some technologies that we need to embrace if we’re going to get emissions reductions down as much as we need to are more obstinate economically. I’m thinking of carbon capture, sequestration, and utilization, for example.

Some environmentalists hate those options, saying it’s just a further lease on life for fossil fuels. But in a world that is still almost 80 percent dependent for its primary energy on coal, oil, and natural gas, we have to recognize that there is going to be quite a lot of fossil fuel burned in the years ahead. What we need to do, given the urgency of the climate change challenge, is make it possible to burn some of that fossil fuel in ways that do not release the resulting carbon dioxide into the atmosphere. But carbon capture and sequestration of that carbon in geologic formations is intrinsically expensive, and we will do it only if it is made more economical with subsidies from the government or regulations. So that is another good characteristic of this new legislation: It would provide a boost to the subsidies for carbon capture and sequestration.

GAZETTE: Might supporting carbon capture and sequestration tell oil companies that they have a place in a future carbon-free economy and bring them on board, or is there little chance of them doing anything other than fighting this tooth and nail?

HOLDREN: The situation with the big oil and gas companies has been changing. Some of them are now very progressive in their approach to this problem. BP, for example, has completely restructured its business plan and is aiming for major reductions in the emissions of their operations, and ultimately the emissions for which they’re responsible. I’m quite sure they’re serious about it. A number of the other big oil and gas companies are moving in that direction. I’ve been saying for decades that we’re not going to solve climate change over the dead bodies of private sector companies. We’re going to solve the problem by finding ways to bring those companies on board in a pathway toward a much more sustainable energy system. Those folks can also do arithmetic — in fact they’re very good at it. They can read the data, and the data on climate change make clear that it’s real, it’s deadly, it’s here. And if we are to at least minimize the worst consequences of climate change, there’s going to have to be a much bigger transition than oil and gas companies until recently were willing to acknowledge. They’re acknowledging it now and trying to figure out how to do it. We should push them along.

GAZETTE: There are tax credits to encourage energy production, including large plants of solar and wind power. Does the legislation do anything to encourage the transmission lines needed to get the power from the places where the wind blows and the sun shines to the big cities?

HOLDREN: The biggest problem with transmission is the permitting. It’s incredibly complex to build a transmission line anywhere because of the number of permits required and the ability of any one of those permit grantors to refuse and to block the project. There is language in this legislation about accelerating and streamlining the permitting processes, and that’s going to be important if it can succeed. But this is a very difficult nut to crack because of people’s affection for due process and for review at various levels. That makes it very difficult to solve the transmission problem. It’s clear that a key to a higher proportion of renewable energy in our national mix is being able to build expanded transmission networks to get the energy from where it’s most economically generated to the places where the largest consumption is taking place. We really do have to master the transmission siting challenges in this country, or we’re cooked.

GAZETTE: What’s left out? How do we get the extra 10 percent emissions reduction that Biden wants?

HOLDREN:  The single most important thing that’s left out — under current political circumstances, there was just no way it was going to get done — is putting a price on carbon emissions across the board. That could be done either with a carbon tax or with a cap-and-trade approach, as was tried initially in the Waxman-Markey bill that failed in the beginning of the Obama administration. Economists of every political stripe, conservatives as well as progressives, will tell you that the single most important and efficient thing we could do to reduce greenhouse gas emissions is to put a price on carbon emissions and let the market figure out the cheapest way to get it done.

In the Obama administration, we did calculations that said if we could have imposed a mere $30 a ton tax on carbon dioxide emissions in 2015, that the reductions by 2025 could have been 32 to 34 percent instead of the 26 to 28 percent target we embraced based on measures that didn’t include a carbon tax or its equivalent. That’s a big difference from a very low carbon tax. If we had a carbon tax of a $75 a ton or $100 a ton, it would be transformative. You could rebate the money to poor and middle-class families. You could spend some of the money to reduce other taxes. The late Dale Jorgenson showed that if you had a carbon tax and if you offset its economic impact by reducing capital gains taxes and income taxes, the economy would be better off after 20 years of having a carbon tax than without it. It’s a very old economic principle that it is more efficient in the societal sense to tax “bads” than to tax “goods.” Capital gains and income are goods and emissions are bad. If we taxed emissions instead of income and made it revenue-neutral overall, as Jorgensen pointed out a couple of decades ago or more, the economy comes out better.

GAZETTE: It seems that the power of that versus the targeted approach is that the tax is economy-wide, so it’s gaining efficiencies in every nook and cranny.

HOLDREN:  And it brings out ingenuity in every nook and cranny. That’s the beauty of it.

Listen to more recordings at The Music Lab.

Global consistencies in infant-directed speech included higher pitch and increased pitch range. Other features that appeared many times in the recordings were contrasting vowels (like emphasizing a long “e” and a short “e” in the same word) and pulse clarity (using a musical voice). When the researchers analyzed singing, they noticed consistent changes in tempo and timbre.

The group trained a machine-learning model to use features like pitch and loudness to guess whether a recording was infant-directed or adult-directed. The model was largely accurate in its determination across the 21 societies and strengthened the case for global consistencies, according to Courtney Hilton, a postdoctoral fellow in the Department of Psychology and a co-first author on the paper.

“Our study provides the strongest test yet of whether there are acoustic regularities in infant-directed vocalizations across cultures,” Hilton said. “It is also really the first to convincingly address this question in both speech and song simultaneously. The consistencies in vocal features offer a really tantalizing clue for a link between infant-care practices and distinctive aspects of our human psychology relating to music and sociality.”

This research builds on earlier studies showing that lullabies and altered speech patterns can have a soothing effect on infants, as well as animal findings demonstrating the clear function of vocalizations such as sounding the alarm for an approaching predator or signaling friendliness and approachability.

“When we were selecting the features with which to look at infant-directed song and speech, we specifically appealed to some baseline bioacoustic principles that can be held constant not only in humans, but across different animals,” said Cody J. Moser, a co-first author on the paper, doctoral candidate at the University of California Merced, and a visiting student in Harvard’s Graduate School of Arts and Sciences. “Things about the way that we alter our pitch toward our infants might have their roots in ways that other animals communicate with each other.”

A secondary part of the study involved members of the public trying to identify whether a vocalization was infant-directed or adult-directed. More than 50,000 English speakers with fluency in 199 languages hailing from 187 countries participated in that portion of the study. Most were accurate in their assessments regardless of their native language or culture. This aspect of the research confirmed that people can pick up on the acoustic markers of infant-directed speech and song, even if they don’t understand the words or cultural references.

How people alter their voices when speaking to infants

Read the full study.

Mehr said this “citizen science” approach was invaluable to the researchers.

“The method we used for this paper — that allowed us to get 50,000 people to tell us if they thought a certain song was for a baby or for an adult — is the same method that we could use to understand the trajectory of music perception ability across a person’s lifespan, or whether people hear structure in music the same way depending on the type of language they speak,” he said. “The citizen science method is an innovation that many psychologists are excited about.”

The findings show striking parallels among cultures when it comes to talking and singing to infants, but Moser warned that the similarities shouldn’t be understood as truly universal features.

“We’re not trying to claim that all societies have infant-directed song or infant-directed speech, but we do now know that when people do tend to sing to their infants or speak to their infants, they tend to do it in the same way,” around the world, he said. “That is very interesting and a new finding.”

“There is, of course, a lot of variation and cultural influence in this practice, but we were also finding commonalities hinting at roots in our shared biology,” added Hilton. “In this study, we were trying to uncover these hidden parts of our minds that shape our everyday behavior and experiences.”

Researchers ran these results through mathematical simulations, which showed that this kind of genetic drift in the two groups could not have happened randomly and lead them to the conclusion that the cause was most likely the result of women not moving around to different islands as much as men.

“Females certainly moved to new islands, but when they did so they were part of joint movements of both females and males” said Reich. “This pattern of leaving the community must have been nearly unique to males in order to explain why genetic differentiation is so much higher in mitochondrial DNA than in the rest of the genome.”

The new study from an interdisciplinary team of geneticists and archaeologists quintuples the body of ancient DNA data from the vast Pacific region called Remote Oceania, the last habitable place on earth to be peopled. It also provides surprising insights into the extraordinarily complex peopling of one of Remote Oceania’s major subregions.

Humans arrived and spread through Australia, New Guinea, the Bismarck Archipelago, and the Solomon Islands beginning 50,000 years ago, but it wasn’t until after 3,500 years ago that they began living in Remote Oceania for the first time after developing the technology to cross open water in long-distance canoes.

This expansion included the region called Micronesia — about 2,000 small islands north of the equator, including Guam, the Marshall Islands, the Caroline Islands, Palau, and the Northern Mariana Islands.

It’s long been a mystery what routes people took to arrive in the region. The discovery that there were five streams of migration into Micronesia helps bring clarity to this mystery and the origins of the people there today.

“These migrations we document with ancient DNA are the key events shaping this region’s unique history,” said Liu, a postdoctoral fellow in Reich’s lab and the study’s lead author. “Some of the findings were very surprising.”

Of the five migrations, three were from East Asia, one from Polynesia, and one involved Papuans from the northern fringes of mainland New Guinea. This represented a new wrinkle for researchers because a prior stream of migration to the southwest Pacific and Central Micronesia came not from the mainland but New Britain, an island chain to the east of it.

The researchers also found that present-day Indigenous people of the Mariana Islands in Micronesia, including Guam and Saipan, derive nearly all their pre-European-contact ancestry from two of the East Asian-associated migrations the researchers detected. It makes them the “only people of the open Pacific who lack ancestry from the New Guinea region,” Liu said.

The researchers consulted with several Indigenous communities in Micronesia for the study. This is the Reich group’s fourth publication of original ancient DNA data from remote Pacific islands.

“It’s important that when we do ancient DNA work, we don’t just write a paper about the population history of a region and then move on,” Reich said. “Each paper raises as many new questions as it answers, and this requires long-term commitment to follow up the initial findings. In the Pacific islands there are so many open questions, so many surprises still to be discovered.”

Supporters for this study included the National Institutes of Health, the John Templeton Foundation, the Allen Discovery Center, and the Howard Hughes Medical Institute.

The Fitzpatrick Scale has been in the open domain for decades and was adopted by the tech industry primarily for use in machine learning and artificial intelligence. Humans must annotate images or videos to train machine-learning algorithms for things like face detections.

“Instead of using this Fitzpatrick Scale, which wasn’t designed to classify diverse skin tones among different groups of people, you would use the Monk Skin Tone Scale, which gives you a more granular understanding,” Monk said. He chose 10 color points to better represent the range of skin tones, opting out of wider range as a larger numbers would begin to create problems with annotation. Extensive fieldwork on skin tone and colorism in the U.S. and Brazil informed the final color selection.

Monk’s collaboration with Google began a few years ago when he was contacted by its Research Responsible AI team, and over time, the work turned to finding an alternative to the Fitzpatrick Scale, the sociologist said. That is when the team learned that Monk had developed a scale of his own.

Google is already using the Monk Skin Tone Scale in its image search. When users search makeup looks, for example, they can refine those searches by skin tone to find more relevant matches. The tech giant also used the scale to update its own Real Tone filters (an initiative it had already launched to deal with the skin-tone problem) on its Pixel cameras so that they can capture a wider, more realistic palette of shades.

“There’ve also been some refinements to search where people have complained that in the past, if you search for certain images, you only get back certain types of people,” Monk said. “One example is if you search for ‘cute babies,’ you might get a completely homogenous set of babies that basically all look white. That’s not a very inclusive experience for people at all.”

Google says the scale will be useful in checking the diversity of results for problems with algorithms. “This has been a longstanding issue, not just at Google, but across the tech industry,” Monk said.

Monk plans to publish research, testing the scale on a diverse group of people. He and the Google team conducted research together to validate the scale in the U.S., which was an extension of his own work.

Developers who do not intentionally design their products to work well across the continuum of skin tones will have products that do not work as well, Monk added. The new scale is open-source so that all companies around the world can make their products more inclusive.

“My hope is that other companies, not just Google, take on the work of auditing and making sure that their products are designed to work equally well across the entire range of skin tones, whether that’s using the Monk Skin Tone Scale or not,” Monk said. “I hope they do that work.”