Beyond EUV lithography, new progress

As is well known, almost all current chips are manufactured using photolithography technology. The most advanced chips are based on EUV lithography, which operates at a wavelength of 13.5 nm and can produce features as small as 13nm (Low NA EUV with a numerical aperture of 0.33), 8nm (High NA EUV with a NA of 0.55), and even 4nm - 5nm (Hyper NA EUV with a NA of 0.7 – 0.75). However, the cost is an extremely complex lithography system with very advanced optical components, costing hundreds of millions of dollars.

Therefore, researchers are looking for better methods, and "Beyond-EUV" has become a research direction for many manufacturers.

According to a report in Cosmos citing a paper published in Nature, researchers at Johns Hopkins University have announced a new chip manufacturing method. This method uses a laser with a wavelength of 6.5nm - 6.7nm (also known as Soft X-ray), which can improve the resolution of photolithography tools to 5nm and below.

Scientists have named their method "Beyond-EUV" (BEUV) - indicating that their technology could replace the industry-standard EUV lithography technology. However, the researchers admit that it will still take them several years to build even an experimental B-EUV tool.

A "Beyond-EUV" Technology

Readers familiar with photolithography technology know that to achieve higher photolithography resolution, one can either increase the numerical aperture (NA) or shorten the wavelength (or do both). The numerical aperture refers to the angle at which light is focused. The larger the angle, the smaller the light spot and the higher the resolution.

Specifically regarding wavelengths, as stated in the article "Development Frontiers and Future Challenges of Lithography Machines for Integrated Circuit Equipment," since the first lithography machine was invented in 1961, lithography machines have evolved from contact-type to proximity-type and then to projection-type. Today, the step-and-scan projection lithography machine is the mainstream. The wavelengths of the light sources used in lithography machines have evolved from the earliest ultraviolet (UV) light sources, such as the g-line (436 nm) and i-line (365 nm) sources, to deep ultraviolet (DUV) light sources represented by KrF (248 nm) and ArF (193 nm), and now to the extreme ultraviolet (EUV) light source with a wavelength of 13.5 nm.

This evolutionary pace is the result of scientists' considerations of these devices and physical principles. Taking DUV and EUV as examples, DUV is generated by excimer lasers. Researchers tested a large number of gases to identify emitters with high emission efficiency. Eventually, they found some good emitters, such as KrF (248 nm), ArF (193 nm), and F2 (157 nm), and developed lithography machines based on them.

On the other hand, EUV is easily absorbed by any material. Even the best mirrors absorb a large portion of it. Even if you find a powerful EUV emitter, you can't build an EUV machine without a good mirror. In other words, the rate-limiting factor for EUV is the mirror, not the light emitter. So you need to test many wavelengths on different materials. The optimal wavelength is 13.5 nm, with a reflectivity of up to 70% on a multilayer Mo/Si mirror. That's why current EUV machines are based on this wavelength.

As for why the 6.7nm wavelength mentioned at the beginning of the article is seen as the next choice, it's because it has the second-highest reflectivity, about half that of 13.5 nm. Theoretically, we can build an EUV machine for this wavelength. However, there are some points to note:

First, although the reflectivity of 6.7 nm is only slightly lower than that of 13.5 nm (61% vs 70%), it should be remembered that EUV light needs to be reflected 11 times before reaching the wafer. This means that any slight drawback will be multiplied by 11. If you calculate it, the transmission efficiency of 6.7 nm is only one-fourth that of 13.5 nm.

Second, for a multilayer mirror to achieve optimal reflectivity, constructive interference must occur between different layers. This means that any slight mismatch between the wavelength and the mirror period will significantly reduce the reflectivity. This effect is more pronounced at shorter wavelengths. You can see that the reflectivity curve for 13.5 nm is like a tower, while the curve for 6.7 nm is like a needle. This poses a severe challenge to the light source.

However, from an industry development perspective, the B-EUV light source is not yet mature. Many researchers have tried various methods to generate radiation at a wavelength of 6.7 nm (for example, plasma generated by a gadolinium laser), but there is currently no industry-standard method. Second, due to the higher photon energy of these shorter wavelengths, they interact poorly with traditional photoresist materials used in chip manufacturing. Third, since light at wavelengths of 6.5nm to 6.7nm is almost completely absorbed by all materials rather than reflected, no multilayer-coated mirrors for such radiation have been produced.

Finally, these photolithography tools must be designed from scratch, and there is currently no ecosystem to support these designs, including components and consumables. In summary, building a B-EUV machine (or a soft X-ray machine?) requires breakthroughs in light sources, projection mirrors, photoresists, and even consumables such as thin films or photomasks.

Several Options for Light Sources

To implement this solution, some manufacturers have made breakthroughs in related light sources and various technologies in the past. For example, at the end of last year, a laboratory in California laid the foundation for the next development of extreme ultraviolet (EUV) lithography technology.

The project, led by the Lawrence Livermore National Laboratory (LLNL), aims to drive the next stage of EUV lithography technology development. The core of the project is the laboratory's self-developed drive system - the large-aperture thulium (BAT) laser. According to the laboratory, the project will test whether the BAT laser can increase the efficiency of EUV light sources to about 10 times that of the current industry-standard carbon dioxide (CO2) lasers.

A startup called Inversion is also using a phenomenon called LWFA (Laser Wakefield Acceleration) to create a compact, high-power light source. According to reports, LWFA uses the interaction between a strong laser pulse and plasma to accelerate electrons to extremely high energies over a very short distance. This process is similar to a surfer riding the wake behind a boat: electrons "surf" on plasma waves and gain energy as they travel.

Inversion expects that the company can use LWFA to accelerate electrons to energies of several GeV over a short distance. These high-energy electrons then pass through a free-electron laser, which uses a magnetic structure to make the electrons emit coherent light of a precise wavelength. LWFA can shrink the traditional particle accelerator used to generate high-energy light by a factor of 1000, down to desktop size - from several kilometers to about one meter.

Based on their estimates, Inversion plans to use its advanced light source to project patterns, just like traditional EUVL. However, the light source can be tuned to a wavelength of 13.5nm or lower, with the next-generation target wavelength being 6.7nm.

In addition, a startup called Lace Lithography AS is developing a lithography technology that uses atoms fired at a surface to define features, with a resolution beyond that of extreme ultraviolet lithography. The company claims on its website: "By using atoms instead of light, we offer chip manufacturers capabilities that are 15 years ahead of current technology, with lower costs and lower energy consumption."

The startup xLight is also using its own technology to break through traditional EUV. The free-electron laser (FEL) is their choice. According to reports, this is essentially a high-power light source that uses electrons to generate light of different wavelengths. A particle accelerator is a system that propels charged particles.

Frankly, free-electron lasers (FELs) and particle accelerators are not new. For years, companies, research institutions, and universities have owned and operated particle accelerators to produce tiny subatomic particles such as protons, neutrons, and quarks. These systems are typically used in physics and other scientific applications. Free-electron lasers (FELs) have existed for a long time.

In xLight's technology, electrons are first injected into a particle accelerator and then enter a free-electron laser (FEL). xLight says: "The FEL uses electrons from the particle accelerator and makes them pass through an undulator with a periodic magnetic field, thereby generating a coherent, high-intensity light beam."

In simple terms, EUV light is generated in the accelerator. Then, the EUV light is transmitted from the particle accelerator device to the wafer fab through a device similar to a photon pipeline. At this point, the EUV light is directed to the sub-wafer fab. Inside the sub-wafer fab, there are various independent systems called "flip stations." According to xLight's video, each flip station is dedicated to an EUV tool in the upper wafer fab. During operation, the EUV light is transmitted to each rotation station in the sub-wafer fab area. Then, each rotation station receives the light and directs it to the EUV system upstairs in the wafer fab. This, in turn, powers the EUV equipment.

In this case, the EUV lithography equipment itself does not contain an LPP light source. Instead, the EUV light is generated in the particle accelerator and then transmitted to the EUV equipment in the wafer fab. This is a simple way to describe a complex process.

Nevertheless, the FEL light source developed by xLight generates four times more power than current LPP devices. xLight says: "By providing up to four times the EUV power, wafer fabs can optimize pattern improvement, increase productivity and yield, thereby generating an additional billions of dollars in revenue per scanner per year and reducing the cost per wafer by approximately 50%. In addition, a single xLight system can support up to 20 ASML systems and has a service life of up to 30 years, reducing capital and operating expenses by more than three times."

Theoretically, xLight's technology can be used for low-numerical-aperture EUV, high-numerical-aperture EUV, and even hyper-numerical-aperture EUV. In terms of research and development, ASML is developing 0.75 hyper-numerical-aperture EUV technology, which aims for the far future.

xLight also has a potential entry point. The company is a member of the Blue-X Alliance. Blue-X, organized by EUV Litho, is proposing the use of 6.7nm wavelength EUV lithography technology. 6.7nm EUV is also a very futuristic technology. Currently, Blue-X has 70 member organizations.

Breakthroughs in Photoresist

Recently, Johns Hopkins University made a new breakthrough in photoresist.

The research team led by Michael Tsapatsis, a professor of chemical and biomolecular engineering at Johns Hopkins University, discovered that metals such as zinc can absorb B-EUV light and emit electrons, which in turn trigger a chemical reaction in an organic compound called imidazole. These reactions make it possible to etch very fine patterns on semiconductor wafers.

Interestingly, although zinc performs poorly under traditional 13.5nm EUV light, it is very effective at shorter wavelengths. This highlights the importance of matching materials with the correct wavelength. To apply these metal-organic compounds to silicon wafers, the researchers developed a technique called chemical liquid deposition (CLD). This method can produce a thin layer called aZIF (amorphous zeolitic imidazolate frameworks). These layered materials are mirror-like and grow at a rate of 1 nanometer per second.

CLD can also quickly test different metal-imidazole combinations, making it easier to find the optimal match for different photolithography wavelengths. Although zinc is very suitable for B-EUV, the team points out that other metals may perform better at different wavelengths, providing flexibility for future chip manufacturing technologies.

"Since different wavelengths interact differently with different elements, a metal that is at a disadvantage at one wavelength may perform well at another," Tsapatsis explains. He reiterates: "Zinc is not very good for extreme ultraviolet radiation, but it is one of the best materials for B-EUV. By adjusting these two components (metal and imidazole), you can change the efficiency of light absorption and the chemical properties of subsequent reactions. This opens the door for us to create new metal-organic pairings. Excitingly, there are at least 10 different metals that can be used in this chemical reaction, as well as hundreds of organic compounds."

Although this research focuses mainly on photoresists for microchip manufacturing, the researchers expect that their findings will also benefit other pattern and thin-film applications, such as sensors and separation membranes.

Although the researchers have not yet solved all the challenges of B-EUV (such as light source power and photomasks), they have overcome one of the most critical bottlenecks: finding a photoresist material that can work under 6-nanometer wavelength light. They invented the CLD process to coat a uniform aZIF thin film on silicon wafers. Through experiments, they showed that certain metals (such as zinc) can absorb soft X-rays and emit electrons, triggering a chemical reaction in the imidazole-based resist.

However, B-EUV still faces many challenges, and the technology has not yet clearly entered the mass market. Nevertheless, the CLD process has a wide range of applications, both in the semiconductor field and non-semiconductor fields.

This article is from the WeChat official account "Semiconductor Industry Observation" (ID: icbank), author: Editorial Department. Republished by 36Kr with permission.