Researchers from the University of Southampton and the National University of Singapore have presented a method for creating two-dimensional heterostructures without using polymers, which are traditionally employed in this process. The results of this study were published in the journal Nature Communications.
The proposed approach replaces adhesive synthetic materials with natural muscovite crystals, which is a type of mica. This allows for a reduction in microscopic contamination that typically degrades the quality of structures composed of layers only a few atoms thick.
This achievement has the potential to increase the accuracy of research using quantum materials and accelerate the development of more reliable electronic components. Surface purity is recognized as a critical factor for the performance of such devices.
The production of two-dimensional materials depends on stacking extremely thin atomic layers. Standard methods for this assembly use synthetic polymers, which leave microscopic residues that alter the electronic behavior of the structures. To overcome this problem, the team used muscovite, a mineral known as mica.
According to the researchers, because muscovite is an inorganic crystal, it provides a more uniform surface and significantly reduces contaminants during the assembly of heterostructures. The work also demonstrated that this substitution allows for the placement of atomic layers with greater precision. Such control is important because some two-dimensional materials acquire completely different properties when stacked at specific angles.
The study's lead author, Marcus Shishkin, a professor of experimental physics at the University of Southampton, emphasized that this level of precision is critical for research in quantum materials. He stated that when two-dimensional materials, such as graphene and hexagonal boron nitride, are stacked into structures with controlled angles between layers, they begin to exhibit entirely new properties.
The researcher explained that small amounts of contamination can distort experimental results, so a cleaner process expands the scope of scientific inquiry. He also noted that the new method simplifies the fabrication of structures and reduces process costs.
In Shishkin's assessment, improving contamination-free fabrication methods is an important step toward creating more efficient nanoelectronic components and, in the future, faster and more reliable microchips.
Alexey Berdyugin, a professor at the National University of Singapore and co-author of the study, noted that creating atom stacks without impurities remains one of the biggest challenges in nanotechnology. He added that since mica is an inorganic crystal rather than a soft polymer, it avoids many contamination issues inherent in traditional methods.
According to Berdyugin, this method provides ultra-pure surfaces, allowing electronic components to operate at maximum performance. The researcher believes this feature could pave the way for both fundamental research and future applications of quantum technologies.
Scientists from the Indian Institute of Science Education and Research (IISER) in Bhopal have discovered a more accurate method for mapping underground systems in one of India's most vital coal regions. By studying microscopic cracks in coal seams in the Jharia basin, researchers created a new mathematical framework explaining the movement of methane and carbon dioxide in the Earth's crust.
This finding provides engineers with a more precise set of tools for extracting clean-burning fuel, as well as, importantly, for safely sequestering greenhouse gases deep underground in the fight against climate change.
The Jharia coal basin, located in the state of Jharkhand, is known for its high-quality coking coal. This region also contains methane in the coal seams (Coal-Bed Methane, CBM)—a form of natural gas trapped within the coal itself. As the world transitions to renewable energy sources, CBM acts as a critical transitional fuel because it burns cleaner than coal or oil.
Furthermore, these same coal seams can function as giant carbon sinks. When carbon dioxide is injected into undeveloped coal layers, the gas can be permanently stored, preventing its release into the atmosphere and global warming.
To implement any of these technologies, engineers need to know precisely the rate at which fluids can move through coal. Unlike a standard pipe, coal is a double-porosity system. This means that gas is stored in the solid matrix of the coal but can only move through a network of tiny natural cracks called cleats. These cleats serve as the highways of the coal seam.
For decades, scientists used a standard formula known as the cubic law to predict this flow. This law assumes that the volume of fluid passing through a crack is proportional to the cube of its width, or aperture. However, the IISER Bhopal team found that this standard rule does not entirely fit the unique, stressed environment of the Jharia basin.
To confirm their findings, researchers collected samples from various conditions across the region, including deep underground mines and extensive open-pit mines, reflecting different depths and geological pressures. Using microscopy and 1D scanning methods, they measured thousands of minute cracks. They discovered that in the Jharia area, gas flow is much more sensitive to crack width than previously thought.
Instead of the third power proposed by the cubic law, they established that the flow actually scales with the fifth power. Specifically, for different types of cleats in the region, the flow followed a power exponent ranging from 4.38 to 5.08.
Researchers believe that the dependence on the fifth power is related to the shape and growth mechanism of the cracks. In Jharia, millions of years of tectonic activity, including folding and crustal faulting, compressed and stretched the coal. This created a complex network where crack width is closely related to its length in a specific scaling relationship. By understanding this relationship, scientists were able to rewrite the flow equations so that they no longer depend on crack length.
In geology, measuring crack length is an extremely difficult task. When engineers drill into the ground to take samples, they obtain only a small cylinder of rock called a core. These cores are often too narrow to show the full length of a crack, leading to insufficient sampling and inaccurate data. By creating a model that requires measuring only the crack width, which is visible even on small samples, researchers have significantly simplified the ability of field engineers to create dynamic fluid flow models without needing inaccessible data.
This research could strengthen India's energy security by making domestic natural gas extraction more predictable and efficient. It also offers a scientific roadmap for large-scale carbon sequestration. By knowing exactly how the plume of carbon dioxide migrates through the coal seam, scientists can ensure that the gas remains where it was placed, turning old coal mines into permanent storage facilities for global carbon waste. As the global community moves towards net-zero emissions, such region-specific scientific knowledge is a vital mechanism that will drive the green economy.
Artificial intelligence (AI) is becoming a much more accessible resource, driven by cost reductions, more efficient models, and greater availability of computing power. While this trend benefits users and corporations, it represents a significant challenge to the financial sustainability of industry leaders like OpenAI and Anthropic.
An analysis published by The Wall Street Journal points out that AI is moving from being a scarce asset toward becoming a commodity. This shift could diminish the competitive advantage of companies currently leading the development of the most sophisticated models.
One of the main reasons for this transformation is the drastic drop in costs for AI capable of performing most daily tasks. This is facilitated by the emergence of lighter models, operable both in the cloud and directly on devices, including solutions from Google, Apple, and Chinese AI companies.
Additionally, China has reinforced its strategy. In a speech delivered in Shanghai on Thursday (16), Chinese President Xi Jinping advocated for maintaining the release of 'open-weights' AI models, allowing for free modification and use. This tactic aims to counterbalance US hegemony in the sector. Chinese models, such as GLM 5.2 and Kimi K3, are narrowing the gap with advanced systems developed by American companies.
Meta has also made considerable advances in developing programming-focused models, demonstrating the ability to compete directly with OpenAI and Anthropic in the lucrative code generation niche. It is expected that the current computational capacity bottleneck will be mitigated by activating new data centers and developing more efficient methods for running AI models. In certain applications, the supply of tokens, the basic unit of consumption, already matches demand.
These advancements are seen as positive for consumers and businesses. About a year ago, Sam Altman, CEO of OpenAI, expressed a desire to make artificial intelligence so cheap that it would be difficult to measure. AI has the potential to boost productivity across various professions and reduce digital friction, rather than simply replacing workers.
However, the scenario raises uncertainties about the financial future of OpenAI and Anthropic, both candidates for initial public offerings. Since they rely on maintaining technological superiority over established giants, they may suffer if AI is treated as a general-purpose technology comparable to electricity or the automobile.
Data from Sensor Tower, cited in the analysis, indicates that in March, ChatGPT accounted for less than 50% of the global consumer user share, considering web and mobile device access. This decrease was mainly attributed to competition from Google Gemini and Anthropic's Claude.
In the corporate sphere, Chinese models have also begun competing with major American systems, showing similar performance in some metrics but at substantially lower costs. The OpenRouter ranking shows that the five most used enterprise models are currently Chinese, and approximately 45% of the tokens monitored by the platform are consumed by these models.
The analysis also highlights the launch of an open-weights model by Thinking Machines Lab, a company led by Mira Murati, former CTO of OpenAI. This system aims to balance performance and operational cost, summarizing the change with the analogy: 'Who needs an AI Ferrari to go to work when an AI Honda Civic is right there?'
Faced with growing competition, OpenAI and Anthropic are increasing their investments in engineers and data center access, even if it affects profitability. Both companies are committing hundreds of billions of dollars to preserve technological leadership. Simultaneously, enterprise clients are scrutinizing the necessary investment in AI more rigorously and which premium models justify their costs.
Another accelerating factor in the transformation of AI into a commodity is the widespread dissemination of knowledge on how to build advanced models. Although companies continue to protect their trade secrets—such as the lawsuit filed by Apple against OpenAI for alleged intellectual property theft—researchers continuously publish scientific studies detailing new advancements. Chinese labs and companies, such as Thinking Machines Lab, also regularly make open-source models available with complete documentation of their creation.
The text also addresses the technique known as 'distillation,' which involves training one model using another as a reference. OpenAI and Anthropic accuse Chinese companies of using this method to create competing systems based on private information. However, the industry itself uses the technique legitimately; large models are often employed to generate smaller and more agile versions.
The new model supporting Apple's updated Siri, for example, was distilled from Google models, according to an agreement between the parties. In a recent essay, Microsoft CEO Satya Nadella described it as 'ironic' that companies training models with internet data and customer information try to block third-party use of distillation. He argued: 'If learning flows in only one direction, economic value converges to the owners of the learning infrastructure, not to the creators of the knowledge itself.'
Previously, the main competitive advantage of OpenAI and Anthropic lay in offering models notably superior to competitors. With the popularization of systems considered 'good enough'—and in some cases, highly advanced—available to both iPhone users and large corporations, these companies need to find new differentiators.
While Google relies on its search engine, Meta possesses its social media base, and Microsoft and Amazon dominate enterprise infrastructure, while Apple controls a vast ecosystem of devices.
Eric Zhao, a professor at the University of Oxford and co-author of a mentioned study, suggests that access to electrical energy may become the primary future competitive advantage for AI companies. As electricity supply becomes constrained and communities oppose the construction of new data centers, energy efficiency will be crucial. He stated that 'frontier labs will need to compete for intelligence per watt.'
OpenAI is expanding its revenue streams, counting over three million corporate clients and developing proprietary hardware to create a direct link with consumers. Meanwhile, Anthropic recently achieved its first profitable quarter and filed an application for a public offering in the Northern Hemisphere autumn (September-December). If the IPO occurs, the company could raise funds to acquire new customers, develop other income sources, or increase data center rentals.
The analysis concludes that the AI market may be robust enough to sustain multiple winning companies, potentially allowing OpenAI or Anthropic to become tech giants. However, the text cites an assessment by the Bank for International Settlements (BIS), according to which the volume of investments in AI already exceeds any peaceful economic expansion cycle, including railway construction and the late 1990s internet bubble.
In this context, critics are growing who doubt whether companies will be able to justify current and future investments. Without a lasting competitive advantage, the companies leading the AI boom could face a considerable downturn in the future.
Researchers announced the first confirmation of an atmosphere on a rocky planet located in the habitable zone of a different star. This advance represents a significant milestone in the search for worlds with the potential to host life.
For decades, scientists have dedicated efforts to identifying planets with conditions favorable for the emergence of life, searching for celestial bodies with characteristics analogous to Earth. Although habitable zones have been mapped in various stellar systems and over six thousand exoplanets cataloged, none until now had gathered the complete set of terrestrial requirements.
For known life to thrive, a planet must possess a rocky structure, exhibit liquid water on its surface—which implies a moderate temperature—and, crucially, be capable of retaining an atmosphere. This gaseous layer is vital as it regulates the climate, protects the surface from radiation, and sustains the conditions necessary for biological existence.
The situation changed with the detection of a rocky planet with an atmosphere in the habitable zone of another star. This achievement was made by researchers from Harvard University and published in the scientific journal Science on July 16. It is important to note that although no life has been found on the planet, the essential elements for its emergence have been identified.
The object studied is the exoplanet LHS 1140 b, which was discovered in 2017. It orbits a red dwarf, located approximately 48 light-years from Earth, within its habitable zone. Red dwarf stars are notorious for emitting intense radiation and destructive flares, but the specific star in this system proved to be less active, allowing the planet to conserve its gaseous layer.
Atmospheric detection was made possible by the use of a theoretical model created by Collin Cherubim, the lead author of the study. This model projected the composition of the planet's atmosphere, suggesting that LHS 1140 b possessed an upper atmosphere rich in helium, which would be gradually escaping into space. To validate this prediction, the team employed the WINERED spectrograph, an instrument that confirmed the presence of this escaping helium around the planet, thus proving atmospheric retention.
David Charbonneau, head of the Department of Astronomy at Harvard University, commented that Collin predicted the helium atmosphere after analyzing known planets, organized observation time with telescopes, collected the data, and that the detection was considered statistically robust.
According to the study, this atmosphere may have been present for more than three billion years. However, there are important distinctions between LHS 1140 b and Earth; the exoplanet is larger, more massive, and colder. Additionally, it completes an orbit in only 25 days and exhibits synchronized rotation, keeping one side always facing its star.
The researchers now plan to investigate the composition of this atmosphere in detail and search for other signs of habitability, such as the possible existence of liquid water on the surface. The model developed by the team will also be applied to the study of other exoplanets.
