Engineers TV

The team’s results appeared recently in the Journal of Advances in Modeling Earth Systems. The study’s MIT co-authors include postdoc Benedikt Barthel Sorensen and Alexis-Tzianni Charalampopoulos SM ’19, PhD ’23, with Shixuan Zhang, Bryce Harrop, and Ruby Leung of the Pacific Northwest National Laboratory in Washington state.

Enormous computing power

Today’s large-scale climate models simulate weather features such as the average temperature, humidity, and precipitation around the world, on a grid-by-grid basis. Running simulations of these models takes enormous computing power, and in order to simulate how weather features will interact and evolve over periods of decades or longer, models average out features every 100km or so.

“It’s a very heavy computation requiring supercomputers,” says Prof Sapsis. “But these models still do not resolve very important processes like clouds or storms, which occur over smaller scales of a kilometre or less.”

To improve the resolution of these coarse climate models, scientists typically have gone under the bonnet to try and fix a model’s underlying dynamical equations, which describe how phenomena in the atmosphere and oceans should physically interact.

“People have tried to dissect into climate model codes that have been developed over the last 20 to 30 years, which is a nightmare, because you can lose a lot of stability in your simulation,” says Prof Sapsis. “What we’re doing is a completely different approach, in that we’re not trying to correct the equations but instead correct the model’s output.”

The team’s new approach takes a model’s output, or simulation, and overlays an algorithm that nudges the simulation towards something that more closely represents real-world conditions.

The algorithm is based on a machine-learning scheme that takes in data, such as past information for temperature and humidity around the world, and learns associations within the data that represent fundamental dynamics among weather features. The algorithm then uses these learnt associations to correct a model’s predictions.

“What we’re doing is trying to correct dynamics, as in how an extreme weather feature, such as the wind speeds during a Hurricane Sandy event, will look like in the coarse model, versus in reality,” says Prof Sapsis.

“The method learns dynamics, and dynamics are universal. Having the correct dynamics eventually leads to correct statistics, for example, frequency of rare extreme events.”

Climate correction

As a first test of their new approach, the team used the machine-learning scheme to correct simulations produced by the Energy Exascale Earth System Model (E3SM), a climate model run by the US Department of Energy, that simulates climate patterns around the world at a resolution of 110km.

The researchers used eight years of past data for temperature, humidity, and wind speed to train their new algorithm, which learnt dynamical associations between the measured weather features and the E3SM model.

They then ran the climate model forward in time for about 36 years and applied the trained algorithm to the model’s simulations. They found that the corrected version produced climate patterns that more closely matched real-world observations from the last 36 years, not used for training.

“We’re not talking about huge differences in absolute terms,” says Prof Sapsis. “An extreme event in the uncorrected simulation might be 105 degrees Fahrenheit, versus 115 degrees with our corrections. But for humans experiencing this, that is a big difference.”

When the team then paired the corrected coarse model with a specific, finer-resolution model of tropical cyclones, they found the approach accurately reproduced the frequency of extreme storms in specific locations around the world.

“We now have a coarse model that can get you the right frequency of events, for the present climate. It’s much more improved,” says Prof Sapsis.

“Once we correct the dynamics, this is a relevant correction, even when you have a different average global temperature, and it can be used for understanding how forest fires, flooding events, and heat waves will look in a future climate. Our ongoing work is focusing on analysing future climate scenarios.”

“The results are particularly impressive as the method shows promising results on E3SM, a state-of-the-art climate model,” says Pedram Hassanzadeh, an associate professor who leads the Climate Extremes Theory and Data group at the University of Chicago and was not involved with the study.

“It would be interesting to see what climate change projections this framework yields once future greenhouse-gas emission scenarios are incorporated.”

Furst has started a company called Helix Carbon to further develop the technology. Former MIT postdoc Gang Fan is the lead author of the paper, which appeared recently in the Journal of the American Chemical Society Au. Other authors include Nathan Corbin PhD ’21, Minju Chung PhD ’23, former MIT postdocs Thomas Gill and Amruta Karbelkar, and Evan Moore ’23.

Breaking down CO₂

Converting carbon dioxide into useful products requires first turning it into carbon monoxide. One way to do this is with electricity, but the amount of energy required for that type of electrocatalysis is prohibitively expensive.

To try to bring down those costs, researchers have tried using electrocatalysts, which can speed up the reaction and reduce the amount of energy that needs to be added to the system. One type of catalyst used for this reaction is a class of molecules known as porphyrins, which contain metals such as iron or cobalt and are similar in structure to the heme molecules that carry oxygen in blood. 

During this type of electrochemical reaction, carbon dioxide is dissolved in water within an electrochemical device, which contains an electrode that drives the reaction. The catalysts are also suspended in the solution. However, this setup isn’t very efficient because the carbon dioxide and the catalysts need to encounter each other at the electrode surface, which doesn’t happen very often.

To make the reaction occur more frequently, which would boost the efficiency of the electrochemical conversion, Prof Furst began working on ways to attach the catalysts to the surface of the electrode. DNA seemed to be the ideal choice for this application.

“DNA is relatively inexpensive, you can modify it chemically, and you can control the interaction between two strands by changing the sequences,” she says. “It’s like a sequence-specific Velcro that has very strong but reversible interactions that you can control.”

To attach single strands of DNA to a carbon electrode, the MIT researchers used two 'chemical handles', one on the DNA and one on the electrode. These handles can be snapped together, forming a permanent bond. A complementary DNA sequence is then attached to the porphyrin catalyst, so that when the catalyst is added to the solution, it will bind reversibly to the DNA that’s already attached to the electrode – just like Velcro.

Once this system is set up, the researchers apply a potential (or bias) to the electrode, and the catalyst uses this energy to convert carbon dioxide in the solution into carbon monoxide. The reaction also generates a small amount of hydrogen gas, from the water. After the catalysts wear out, they can be released from the surface by heating the system to break the reversible bonds between the two DNA strands, and replaced with new ones.

An efficient reaction

Using this approach, the researchers were able to boost the Faradaic efficiency of the reaction to 100%, meaning that all of the electrical energy that goes into the system goes directly into the chemical reactions, with no energy wasted. When the catalysts are not tethered by DNA, the Faradaic efficiency is only about 40%.

This technology could be scaled up for industrial use fairly easily, Furst says, because the carbon electrodes the researchers used are much less expensive than conventional metal electrodes. The catalysts are also inexpensive, as they don’t contain any precious metals, and only a small concentration of the catalyst is needed on the electrode surface.

By swapping in different catalysts, the researchers plan to try making other products such as methanol and ethanol using this approach. Helix Carbon, the company started by Furst, is also working on further developing the technology for potential commercial use.

Such operators also have a harder time responding to environmental regulations like the Environmental Protection Agency’s new methane guidelines, which seek to dramatically reduce emissions of the potent greenhouse gas in the industry over the next few years.

“These operators don’t want to be releasing methane,” says Mannai. “Additionally, when gas gets into the pumping equipment, it leads to premature failures. We can detect gas and slow the pump down to prevent it. It’s the best of both worlds: the operators benefit because their machines are working better, saving them money while also giving them a smaller environmental footprint with fewer spills and methane leaks.”

Cutting-edge technology

Mannai learnt about the cutting-edge technology used in the space and aviation industries as he pursued his master’s degree at the Gas Turbine Laboratory in MIT’s Department of Aeronautics and Astronautics. Then, during his PhD at MIT, he worked with an oil services company and discovered the oil and gas industry was still relying on decades-old technologies and equipment.

“When I first travelled to the field, I could not believe how old-school the actual operations were,” says Mannai, who has previously worked in rocket engine and turbine factories. “A lot of oil wells have to be adjusted by feel and rules of thumb. The operators have been let down by industrial automation and data companies.”

Monitoring oil wells for problems typically requires someone in a pickup truck to drive hundreds of miles between wells looking for obvious issues, says Mannai. The sensors that are deployed are expensive and difficult to replace. Over time, they’re also often damaged in the field to the point of being unusable, forcing technicians to make educated guesses about the status of each well.

“We often see that equipment unplugged or programmed incorrectly because it is incredibly overcomplicated and ill-designed for the reality of the field,” says Mannai. “Workers on the ground often have to rip it out and bypass the control system to pump by hand. That’s how you end up with so many spills and wells pumping at suboptimal levels.”

To build a better oil field monitoring system, Mannai received support from the MIT Sandbox Innovation Fund and the Venture Mentoring Service (VMS).

He also participated in the delta V summer accelerator at the Martin Trust Center for MIT Entrepreneurship, the fuse program during IAP, and the MIT I-Corps program, and took a number of classes at the MIT Sloan School of Management. In 2019, Amplified Industries – which operated under the name Acoustic Wells until recently – won the MIT $100K Entrepreneurship competition.

“My approach was to sign up to every possible entrepreneurship related program and to leverage every MIT resource I possibly could,” says Mannai. “MIT was amazing for us.”

Mannai officially launched the company after his postdoc at MIT, and Amplified raised its first round of funding in early 2020. That year, Amplified’s small team moved into the Greentown Labs startup incubator in Somerville.

Huge challenge

Mannai says building the company’s battery-powered, low-cost sensors was a huge challenge. The sensors run machine-learning inference models and their batteries last for 10 years. They also had to be able to handle extreme conditions, from the scorching hot New Mexico desert to the swamps of Louisiana and the freezing cold winters in North Dakota.

“We build very rugged, resilient hardware; it’s a must in those environments” says Mannai. “But it’s also very simple to deploy, so if a device does break, it’s like changing a lightbulb: we ship them a new one and it takes them a couple of minutes to swap it out.”

Customers equip each well with four or five of Amplified’s sensors, which attach to the well’s cables and pipes to measure variables like tension, pressure, and amps.

Vast amounts of data are then sent to Amplified’s cloud and processed by their analytics engine. Signal processing methods and AI models are used to diagnose problems and control the equipment in real-time, while generating notifications for the operators when something goes wrong. Operators can then remotely adjust the well or shut it down.

“That’s where AI is important, because if you just record everything and put it in a giant dashboard, you create way more work for people,” says Mannai. “The critical part is the ability to process and understand this newly recorded data and make it readily usable in the real world.”

Amplified’s dashboard is customised for different people in the company, so field technicians can quickly respond to problems and managers or owners can get a high-level view of how everything is running.

Mannai says often when Amplified’s sensors are installed, they’ll immediately start detecting problems that were unknown to engineers and technicians in the field. To date, Amplified has prevented hundreds of thousands of gallons worth of brine water spills, which are particularly damaging to surrounding vegetation because of their high salt and sulphur content.

Preventing those spills is only part of Amplified’s positive environmental impact; the company is now turning its attention toward the detection of methane leaks.

Helping a changing industry

The EPA’s proposed new Waste Emissions Charge for oil and gas companies would start at $900 per metric ton of reported methane emissions in 2024 and increase to $1,500 per metric ton in 2026 and beyond.

Mannai says Amplified is well positioned to help companies comply with the new rules. Its equipment has already showed it can detect various kinds of leaks across the field, purely based on analytics of existing data.

“Detecting methane leaks typically requires someone to walk around every valve and piece of piping with a thermal camera or sniffer, but these operators often have thousands of valves and hundreds of miles of pipes,” says Mannai.

“What we see in the field is that a lot of times people don’t know where the pipes are because oil wells change owners so frequently, or they will miss an intermittent leak.”

Ultimately Mannai believes a strong data backend and modernised sensing equipment will become the backbone of the industry, and is a necessary prerequisite to both improving efficiency and cleaning up the industry.

“We’re selling a service that ensures your equipment is working optimally all the time,” says Mannai. “That means a lot fewer fines from the EPA, but it also means better-performing equipment. There’s a mindset change happening across the industry, and we’re helping make that transition as easy and affordable as possible.”

Led by Ju Li, the Battelle Energy Alliance Professor in Nuclear Engineering and professor of materials science and engineering, the team described a framework for understanding precisely how changing the elastic strains on a material can fine-tune properties like thermal and electrical conductivity. The work is described in an open-access paper published in PNAS.

“For the first time, by using machine learning, we’ve been able to delineate the complete six-dimensional boundary of ideal strength, which is the upper limit to elastic strain engineering, and create a map for these electronic and phononic properties,” says Prof Li. “We can now use this approach to explore many other materials. Traditionally, people create new materials by changing the chemistry.”

“For example, with a ternary alloy, you can change the percentage of two elements, so you have two degrees of freedom,” he continues. “What we’ve shown is that diamond, with just one element, is equivalent to a six-component alloy, because you have six degrees of elastic strain freedom you can tune independently.”

Small strains, big material benefits

The paper builds on a foundation laid as far back as the 1980s, when researchers first discovered that the performance of semiconductor materials doubled when a small – a mere 1% – elastic strain was applied to the material.

While that discovery was quickly commercialised by the semiconductor industry and today is used to increase the performance of microchips in everything from laptops to cellphones, that level of strain is very small compared to what we can achieve now, says Subra Suresh, the Vannevar Bush Professor of Engineering Emeritus.

In a 2018 Science paper, Prof Suresh, Dao, and colleagues demonstrated that 1% strain was just the tip of the iceberg.

As part of a 2018 study, Prof Suresh and colleagues demonstrated for the first time that diamond nanoneedles could withstand elastic strains of as much as 9% and still return to their original state. Later on, several groups independently confirmed that microscale diamond can indeed elastically deform by approximately 7% in tension reversibly.

“Once we showed we could bend nanoscale diamonds and create strains on the order of 9% or 10%, the question was, what do you do with it,” says Prof Suresh.

“It turns out diamond is a very good semiconductor material … and one of our questions was, if we can mechanically strain diamond, can we reduce the band gap from 5.6 electron-volts to two or three? Or can we get it all the way down to zero, where it begins to conduct like a metal?”

To answer those questions, the team first turned to machine learning in an effort to get a more precise picture of exactly how strain altered material properties.

“Strain is a big space,” says Prof Li. “You can have tensile strain, you can have shear strain in multiple directions, so it’s a six-dimensional space, and the phonon band is three-dimensional, so in total there are nine tunable parameters. So, we’re using machine learning, for the first time, to create a complete map for navigating the electronic and phononic properties and identify the boundaries.”

Armed with that map, the team subsequently demonstrated how strain could be used to dramatically alter diamond’s semiconductor properties.

“Diamond is like the Mt Everest of electronic materials,” says Prof Li, “because it has very high thermal conductivity, very high dielectric breakdown strengths, a very big carrier mobility. What we have shown is we can controllably squish Mt Everest down … so we show that by strain engineering you can either improve diamond’s thermal conductivity by a factor of two, or make it much worse by a factor of 20.”

New map, new applications

Going forward, the findings could be used to explore a host of exotic material properties, says Prof Li, from dramatically reduced thermal conductivity to superconductivity.

“Experimentally, these properties are already accessible with nanoneedles and even microbridges,” he says. “And we have seen exotic properties, like reducing diamond’s [thermal conductivity] to only a few hundred watts per meter-Kelvin. Recently, people have shown that you can produce room-temperature superconductors with hydrides if you squeeze them to a few hundred gigapascals, so we have found all kinds of exotic behaviour once we have the map.”

The results could also influence the design of next-generation computer chips capable of running much faster and cooler than today’s processors, as well as quantum sensors and communication devices. As the semiconductor manufacturing industry moves to denser and denser architectures, Prof Suresh says the ability to tune a material’s thermal conductivity will be particularly important for heat dissipation.

While the paper could inform the design of future generations of microchips, Zhe Shi, a postdoc in Prof Li’s lab and first author of the paper, says more work will be needed before those chips find their way into the average laptop or cellphone.

“We know that 1% strain can give you an order of magnitude increase in the clock speed of your CPU,” says Shi. “There are a lot of manufacturing and device problems that need to be solved in order for this to become realistic, but I think it’s definitely a great start. It’s an exciting beginning to what could lead to significant strides in technology.”