There is a strong demand for new chip materials!

The development approach combines old and new technologies, but bringing any new material into mass production is a complex process.

Materials suppliers are under intense pressure to improve power, performance, scaling and cost issues, which require a long timeline from synthesis to development to high-volume production in factories. Advances in machine learning can help provide a broad field of candidates, which engineers can then narrow down to potential uses.

When building a standard logic semiconductor chip, the main materials are obvious - silicon, silicon nitride and oxide, and metals. They are enclosed in a package made of resin and metal sheets. But this brief overview seriously underestimates the number of other supporting materials required to manufacture the finished device. Although the chip industry has been working on it for many years, the number of new materials required to advance the wafer and packaging processes seems endless.

Some of these materials are temporary and will be removed during processing. Others provide replacements for silicon nitride or silicon dioxide to achieve better etch selectivity. Still others act as hard masks, adhesives, or many other mundane roles. All are critical, but depending on the goals of the new material, they can take years to develop. Looking ahead is critical to ensuring that new silicon or packaging process development proceeds smoothly to production.

The need for a new material may become obvious when it becomes the basis for a new process or process variant, but there are many other reasons why a different material is needed. “We worked with a customer who needed to qualify a new photoresist because the original photoresist was no longer available,” said David Park, vice president of marketing at Tignis. Regardless of the motivation, the development of a new material can be a lengthy process.

Thinned wafers for 3D stacking in advanced packaging. Source: Brewer Science

Companies like Brewer Science provide the materials needed by the semiconductor industry. Fabs and semiconductor equipment manufacturers rely on these companies to implement new processes that require different inputs. While the uses of different materials may be very different, their development has many things in common.

Ask different material suppliers what a new material is and you'll get a variety of answers. The most obvious answer is to synthesize a material that didn't exist before. This is actually relatively uncommon. More common is to find new ways to create existing substances using more efficient or more effective processes.

Some of the materials being developed are improvements on existing ones to suit new requirements. Others may represent a new combination of elements rather than a new molecule. The range of considerations has expanded dramatically. “Twenty years ago, we were only thinking about seven or 10 elements from the periodic table for use in chips,” says Anand Nambiar, chief commercial officer for EMD Electronics, the electronics business unit of Merck KGaA, Darmstadt, Germany, in the U.S. and Canada. “Today, there are at least 70 or 75 elements in high-volume production or under development in some form.”

Historically, most of these materials (precursors, photoresists, masks) have been organic. But metal oxides are starting to play a bigger role in patterning and etching. “Traditionally, Brewer has designed organic materials and added silicon materials for hard mask applications over the last 20 years,” explains James Lamb, a researcher at Brewer Science. “But we’ve also added metal hard masks, so we’re now moving into the inorganic space.”

Since many materials already exist, it is not common to patent new substances themselves. What is important is the method of making them, and process patents are more common. But the formulation, the material and how it is applied form a whole and must be considered together. In some cases, the same molecule may be used in different places or layers, but at different thicknesses. Some people also consider them different materials, given the formulation.

Important material properties begin with how well it performs in the application, but don’t end there. Access, deposition, side reactions, and many other considerations can affect whether a substance is suitable for high-volume manufacturing.

Conversations with materials companies reveal the kinds of materials being worked on today. Brewer Science, for example, has been improving package underfill materials—materials that provide physical stability when building packaged chips. They must be electrically inert. “Our goal is to make materials that don’t need to short,” said Rama Puligadda, chief technology officer at Brewer Science. “They can’t be electrically conductive. They have to be thermally conductive.”

Spin-on carbon materials are able to improve patterning when using thin photoresists. The key requirement here is that they must remain stable at high temperatures, such as those during the deposition process. "These things can't outgas, because it would cause the chemical vapor deposition (CVD) material or atomic layer deposition (ALD) material to pop and crack at high temperatures," Lamb explained.

Almost by definition, new battery chemistries require new materials. “In some way, shape or form, the range of a car or the fast charging depends on materials,” said Puneet Sinha, senior director, global head of batteries at Siemens Digital Industries Software.

Atomera uses a fairly common material - oxygen. What's new is how it's applied and used. For example, during epitaxial silicon growth, the wafer is exposed to oxygen. Oxygen is not sufficient to form SiO2, as two oxygen atoms are required for each silicon atom. Instead, silicon atoms at the surface have dangling bonds that can connect to a single oxygen atom.

“We take oxygen atoms and dope them in a monolayer,” said Scott Bibaud, CEO of Atomera. “But it’s not an oxide. It’s a partial monolayer of oxygen. So you have a perfect silicon lattice underneath and a distorted field in the middle where you have the oxygen atoms. But on top of that, you can continue to grow perfect silicon.”

This stack has several uses. Because the bond has some mechanical rotational freedom, it can act as a transition layer between two materials with different coefficients of thermal expansion (CTE), helping to prevent cracking or other reliability issues. For example, gallium nitride (GaN) is grown on silicon and is prone to cracking when it cools. Atomera says its oxygen layer can relieve stress caused by the mismatch in stress and CTE between GaN and silicon. In the transistor gate, some of the oxygen can float up to clean up the gate oxide transition. It can also act as an impurity getter. It can reduce transistor variability.

Another key materials development involves the manufacture of EUV pellicles (the material that protects extreme ultraviolet photomasks) and lens coatings. Canatu (named after carbon nanotubes, or CNTs) makes CNT-based pellicles. Some are used as pellicles for extreme ultraviolet (EUV) lithography. Others are used as lens heaters, forming thin films on camera or lidar lenses. Given the large surface area of ​​the materials, the company is also evaluating their usefulness in sensitive sensors.

In EUV pellicles, the process is a major development. Carbon nanotubes are not new, but it has been difficult to produce them on a large scale. Canatu said it has developed a floating catalyst CVD process that can make carbon nanotubes in just two steps. In comparison, the company said, competitors' processes require nine steps, which produces a large amount of carbon nanotube powder that must then be filtered. Traditional metal silicides currently used in EUV pellicles require more than 100 steps.

“We have two different types of reactors,” says Juha Kokkonen, CEO of Canatu. “We have one reactor optimized for the semiconductor sector, where the most important elements are clean processes and a homogeneous, independent network. Durability is required (because the film is subject to high G-forces when the scanner moves), high UV transmittance and temperature tolerance of up to 1500°C. The lens heater requires a different set of features. “For the camera and lidar sensors, we are optimizing the electrical properties, such as high conductivity and slightly larger volume.”

Entegris has created a material that solves the challenge of determining the work function of nanosheet gate-all-around transistors. “At the transistor level, you tune the work function by adjusting the thickness of the metal on top of the high-k dielectric, and that thickness can be up to 150Å,” explains Paul Besser, senior director of advanced technology programs at Entegris. “But the spacing between nanosheets is only 100Å.” Instead, Entegris created dipole shifters (dopants that change the band structure) to perform this function.

Another material Entegris is working on is an etch stop layer for building power delivery on the back side of the wafer. Wafers must be thinned from 700 mm to 50 mm and below, but chemical mechanical polishing (CMP) takes too long. Most of the material removal is done by grinding, and finally CMP and plasma/wet etching are used for cleaning. To prevent etching from going all the way through the wafer, a transparent etch stop layer is necessary. Entegris is working on SiGe layers within the wafer to provide an etch stop layer.

The challenge for Entegris’ process is using solid precursors, rather than the more common liquid precursors. Solids sublimate—go directly from solid to gas—and the gas must make its way all the way to the wafer without any side reactions. “If you volatilize at a certain temperature—let’s say 150 degrees—if the temperature anywhere along the path to the wafer is less than 150 degrees, it will deposit,” Besser said.

To create a new material—or to develop a new application or process for an existing one—there are several options. Assuming no unexpected problems arise, tweaking an existing material is the quickest route. If that’s not possible, then an entirely new synthetic route must be discovered, which might (but usually doesn’t) lead to a new material. “If [the desired properties] are within the range of modification, then we can do that,” explains Hidenori Abe, executive director of Resonac’s electronics business. “But if they need a two-, three-, or five-fold improvement in performance, then we need to design it from scratch.”

The path chosen may depend on how robust the material or process is. “In deposition precursors, you’re constantly trying to find newer molecules,” Nambiar says. “That’s compared to photoresist, which is a formulated material. You don’t change the main ingredients, but you try to tweak some small additive to make it slightly better.”

“We almost always start with what we know. That’s the quickest transfer,” said Brewer Science’s Lamb. “But in our process, we don’t usually do that. We also look for alternative platforms/chemistries unless someone wants a thicker or thinner version of an existing material.”

More typical is tuning a specific material property, such as heat resistance or viscosity. Gases are slightly different because they are individual molecules, and development efforts primarily address the logistical challenges of delivering uniform, predictable amounts of gas to the desired surface, assuming those molecules already exist.

Customers drive development, but they don’t always know exactly what they want. “Sometimes they come to us and say, ‘We want this material with a ligand. Can you make it?’ Or, they have a specification,” Abe says. The properties in the specification form the requirements for the new material without specifying the material directly.

Development is often a collaborative effort. Materials don’t exist in isolation. Instead, they interact with the equipment used to apply them. New processes often involve both equipment changes and new materials. Chipmakers have traditionally worked separately with equipment companies and material suppliers, but they’ve found that integrating materials with equipment late in the process often ends up requiring reworking of the materials.

“We used to bring in our own material, but sometimes we couldn’t get the volume right,” explains Entegris’ Besser. “We had to go back to the equipment supplier and ask them, ‘Can you do this volume?’ That left us with no equipment available.”

More often now, fab customers will work directly with equipment suppliers, and equipment makers will work with materials companies so that the equipment/materials combination works well. “So Lam or Applied Materials will develop the material with us, and then they both will introduce it to the customer,” Nambiar said. The customer then chooses between the equipment/materials options.

Materials development is not a quick process. “Most new materials take years to develop and can take years to launch,” Nambiar said. Canatu’s EUV pellicle took seven years to develop. “We listen to what [customers] might want to do in five or ten years, and then we work together to do a proof of concept,” Abe said.

Any new project must include thorough preliminary research to understand what patents may already exist. Although materials companies may want to be self-sufficient, they may encounter material needs that are already protected by patents. If a company can find a synthesis formula that bypasses an existing patent, it can develop it on its own. If this is not possible or practical, then it may need to work with other companies to license the process or obtain precursors from them.

The materials development process involves a combination of experimentation and simulation. Most materials are not designed directly, but rather derived through measurement and data. “I still think of [material development] as being more empirical,” Lamb muses.

Atomera, for example, developed its technology using ab initio simulation, starting from first principles. Other projects rely on empirical data and carefully designed experiments to determine the best recipe. Merck/EMD can slice wafers and place different film stacks on different parts of the wafer to reduce costs and speed up learning cycles.

Machine learning (ML) helps build and run experiments. “We use many different models to feed the ML algorithms that run DoE [design of experiments],” Nambiar says. “The ML engine gives us hundreds or thousands of possibilities, a few of which are real possibilities. When chemists look at it, they say those are nonsense, and these are real possibilities. It would take us 100 years to come up with these choices.”

Automation helps, too. “With continuous flow, you can run 100 different reactions in an hour or two, and with online analysis, it feeds all of that back into the system,” Lamb says. People can automatically use the results to guide further exploration.

Data mining is especially valuable for companies with a long history of development. When researching new requirements, it’s useful to have access to all the data, even from projects that didn’t succeed. Some discontinued projects may be given a new lease of life. “We’ve really modernized our manufacturing and pilot lines to better document what’s going on in the reaction process,” Lamb says. The company has a data mining tool “… where we can take all the information that’s been generated from all the batch work, testing, and evaluations, and mine that for direction for our current set of materials.”

Not all companies are willing to use all substances. Two factors in particular come into play: safety and sustainability.

Making semiconductors has always involved some hazardous chemicals. College classes warned of the dangers decades ago, for example, of the hydrofluoric acid used as an etchant in an era when humans moved wafers and cassettes between workstations by hand.

Some materials may have changed since then, but the risk remains. “The industry is introducing metal organic resists, which are tin-based molecules,” EMD’s Nambiar points to as another example. “Depending on the type of tin oxide you’re talking about, the toxicity can vary.” While companies typically try to focus on nontoxic material inputs and outputs whenever possible, that’s not always possible.

Pyrophoric substances such as silane are another example of hazardous materials – not because they are toxic in themselves (although they can be), but because they can spontaneously ignite when they come into contact with air or water. This presents a host of logistical challenges, as safety considerations must be addressed while the material is on site and in storage, as well as throughout its transportation and storage.

As a result, some companies try to avoid shipping certain substances with these hazardous properties. “We don’t deal with pyrophoric substances,” Besser said. He explained that when working with aluminum, for example, “… you can deal with TMA [trimethylaluminum], which is a pyrophoric substance, and we don’t want to deal with that because of the health risks.”

Sustainability also involves environmental, geopolitical and human rights considerations. The mining of some raw materials may face any or all of these challenges. For example, lithium and cobalt mining have issues with working conditions for miners. China controls many rare earth resources, making them more politically risky.

“We have no conflict minerals at all,” Besser said. “We avoid geopolitical areas. All of our suppliers sign a business agreement promising not to use any conflict minerals when supplying us. We turn down any business opportunities for the safety of our employees and supply chain.”

Other materials are being phased out for environmental reasons. “We are still using PFAS [per- and polyfluoroalkyl substances] in some of our materials,” Nambiar says. “Our customers are asking us to remove these PFAS.”

Once a new material or synthetic pathway is ready for production, something needs to happen to convert it from lab scale to commercial scale. “It’s one thing to do research, and it’s another to scale it up and deliver it safely at the right cost,” Nambiar says.

Others agree. “The scale-up from flask to production is one of the biggest risk factors you face,” Lamb notes. “You can do most of it in the lab and it’s not too expensive. But once you start scaling up a pilot line, we have to build a new pipeline and manufacturing capabilities to support it.”

This process of scaling up may not be simple. The reactors and other infrastructure that make small batches may not work well for large-scale production. For example, heating a volume of gas to a uniform temperature may be more difficult in a large reactor with external heating elements than in a small reactor, because the center of the reactor may be more difficult to maintain a certain temperature. New analytical methods may also be needed to ensure the quality of a material once it is produced in large quantities.

It may not be possible to plan for scale-up at the beginning of a project because the necessary processes and equipment may not yet be known. But as experiments begin to identify solutions, full-scale production must be considered as soon as possible, as this may require new equipment or relationships with other suppliers. Laboratory equipment and production equipment are often made by different companies, and new equipment must be thoroughly vetted. "We have to cross-check construction materials because we are always dealing with ionic contamination," Lamb says.

Availability of raw materials may also be a consideration. “These raw materials may or may not be close to semiconductor grade,” Lamb noted. “If you need the properties they provide, they may need to be rigorously purified. This takes a lot of time and adds cost.”

The need to scale up indicates success in developing new materials. Some projects may never get that far, due to changes in the market rather than failures of the materials themselves. Before EUV lithography went into production, there was an effort to develop an interim lithography wavelength of 157nm. That technology ultimately gave way to EUV, and 157nm never went into production. Even the development of EUV, which ultimately succeeded, involved many alternative approaches, most of which did not succeed. Constant change is a challenge for those materials being developed.

Ultimately, materials development is a process that relies on humans to provide guidance, knowledge, and expertise. The field is full of PhDs working on arcane concepts, but having a PhD alone isn’t enough. “We try to hire a lot of chemists with PhDs, and make sure they’re not all lithography [experts],” Lamb says. “So we have inorganic chemists, organometallic chemists, pure synthetic chemists, and materials scientists. You get a lot of cross-pollination of ideas, which is really helpful.”

The process also benefits from what’s known as “tribal knowledge.” These experimental databases are essential, but so is the team’s experience mining them. Years of materials development develop intuitions and instincts about potentially useful directions that a more naive approach might miss.

Given the ongoing challenges of the post-Denard scaling era, clever ideas will increasingly be needed to maintain advances in semiconductor cost, performance, and power consumption. Since many of these ideas explore uncharted territory, it is a safe bet that new materials will be necessary to bring them to production. The demand for companies and engineers with materials skills should continue for a long time.

Today is the 3893rd content shared by "Semiconductor Industry Observer" for you, welcome to follow.

★ Important report on EUV lithography machine released by the United States

Silicon carbide "surge": catching up, involution, and substitution

★ Chip giants all want to “kill” engineers!

Apple , playing with advanced packaging

★ A historic moment for GPU!

★ Continental Group, developing 7nm chips

★ Zhang Zhongmou's latest interview: China will find a way to fight back

★ EUV lithography’s new “savior”