कृत्रिम बुद्धिमत्ता
एआई ने अगली पीढ़ी की बैटरियों के लिए नई सामग्री की खोज की
लिथियम बनाम बाकी
Lithium-ion batteries have so far dominated the electrification landscape, in large part due to lithium atoms’ unique electrical properties. Simply put, lithium, being the 3rd lightest element in the periodic table, is the most powerful one when it comes to carrying charges with a single electron.
स्रोत: Medium
हालाँकि, लिथियम महंगा है, जिससे वैकल्पिक बैटरी रसायन विज्ञान संभावित रूप से आर्थिक रूप से आकर्षक बनते हैं। विशेष रूप से, सोडियम-आयन बैटरियों ने इस कारण से लोकप्रियता हासिल की है।
ऐसा प्रतीत होता है कि एक अन्य डिज़ाइन पहले की अपेक्षा अधिक संभावनाएँ रख सकता है: मल्टीवैलेंट-आयन बैटरियां। ये धातु आयनों का उपयोग करती हैं जो एक साथ एक से अधिक इलेक्ट्रॉन ले जा सकते हैं, और लिथियम-आयन बैटरियों की तुलना में अधिक लागत‑प्रभावी हो सकती हैं।
The recent breakthrough was achieved by using AI to test millions of combinations for the battery materials. This discovery was done by researchers at the New Jersey Institute of Technology (NJIT) and the Rensselaer Polytechnic Institute (RPI) (USA). They published their results in Cell Reports Physical Science1, under the title ”Generative AI for discovering porous oxide materials for next-generation energy storage”.
आयन बैटरियों के कई प्रकार
यदि लिथियम-आयन ने अपनी ऊर्जा घनत्व के कारण छोटे इलेक्ट्रॉनिक्स और प्रारंभिक ईवी डिज़ाइनों पर हावी हो गया, तो कई अन्य धातु आयनों को उसी सिद्धांत के अनुसार उपयोग किया जा सकता है।
जैसा कि चर्चा की गई, सोडियम-आयन वर्तमान में एक लोकप्रिय विकल्प है, जो सस्ते ईवी मॉडलों के लिए बड़े पैमाने पर निर्मित हो रहा है।
एक अन्य विकल्प है मैग्नीशियम, कैल्शियम, एल्युमिनियम या जिंक का उपयोग करना, जो सभी मल्टीवैलेंट आयन हैं। इसका अर्थ है कि वे दो या यहाँ तक कि तीन सकारात्मक चार्ज ले जाते हैं।
हालाँकि, बड़े चार्ज के साथ बड़े आकार भी आते हैं। बड़े परमाणु आकार मल्टीवैलेंट आयनों को बैटरी सामग्री में कुशलता से समायोजित करना चुनौतीपूर्ण बनाते हैं, जिससे बैटरी घनत्व बहुत कम हो जाता है और व्यावसायिक रूप से व्यवहार्य नहीं रहता।
कम से कम, यह लिथियम या सोडियम आयनों के लिए विकसित पारंपरिक बैटरी सामग्री के साथ सत्य था। लेकिन वे उपयोग की जा सकने वाली एकमात्र संभावित बैटरी सामग्री नहीं हैं। कई अन्य क्रिस्टलीय संरचनाएँ बनाई जा सकती हैं जो उन आयनों को समाहित कर सकें जिनकी गति विद्युत चार्ज ले जाती है।
“सबसे बड़ी बाधाओं में से एक आशाजनक बैटरी रसायनों की कमी नहीं थी — बल्कि लाखों सामग्री संयोजनों का परीक्षण करने की पूर्ण असंभवता थी,”
Professor Dibakar Datta – New Jersey Institute of Technology (NJIT)
AI अनुसंधान में सहायता
एक शक्तिशाली सहायक
Human minds are not the best at handling any data set where the numbers are going toward the millions. But AIs are excellent at it.
It is a growing trend of researchers, especially in material sciences or biotech, using AI technology to help identify the most promising ideas, before analyzing and testing them more rigorously.
“We turned to generative AI as a fast, systematic way to sift through that vast landscape and spot the few structures that could truly make multivalent batteries practical.
Professor Dibakar Datta – New Jersey Institute of Technology (NJIT)
Previously, a computing model relying solely on physics would have been unable to handle the extremely complex calculation required to model a new type of crystal structure.
But new types of AI, based on machine learning and neural networks, are more able to “guess” the general property of a material without formal mathematical calculation of the physics behind.
स्रोत: Cell
The researchers developed a system relying on two different types of AI at the same time, one knowledgeable about crystals, and one LLM (Large Language Model), the same technological base behind ChatGPT.
स्रोत: Cell
क्रिस्टल डिफ्यूजन वैरिएशनल ऑटोएन्कोडर (CDVAE)
The CDVAE model generated 10,000 structures, which were subjected to a series of precise screening and validation steps to ensure they met the necessary standards.
For example, it checked that the distance between atom pairs was large enough, or the charge neutrality of the system.
This method generated 42 structures potentially usable for battery materials.
Of these, 21 structures matched existing entries in the database but offered new configurations with differences in stoichiometry, lattice parameters, or space groups. The remaining 21 structures were entirely novel.
स्रोत: Cell
So it created both new versions of existing material that were previously unknown and have entirely new potential battery material.
LLM
The researchers then used Meta’s (FB ) Llama-3.1-8B, specially calibrated and tailored for generating crystal structures.
स्रोत: Cell
This generated 10,000+ crystal structures, of which 1,087 structures remained after checking for structural integrity. Using the same filters as with CDVAE, this resulted in 13 potential candidates, of which the 5 most stable structures were chosen.
स्रोत: Cell
स्क्रॉल करने के लिए स्वाइप करें →
| मॉडल | प्रारंभिक संरचनाएँ | फ़िल्टर‑के‑बाद उम्मीदवार | अंतिम स्थिर सामग्री |
|---|---|---|---|
| CDVAE | 10,000 | 42 | 21 वैरिएंट + 21 नया |
| LLM (Llama-3.1-8B) | 10,000+ | 13 | 5 सबसे स्थिर चुने गए |
AI की खोजों को चुनौती देना
The researchers used a mathematical testing method called “DFT relaxation”, calculating the material free energy (linked to stability), to check the quality of the material found.
It quickly appears that the LLM-generated crystalline materials were generally much better and more stable than those generated with CDVAE.
स्रोत: Cell
“Our AI tools dramatically accelerated the discovery process, which uncovered five entirely new porous transition metal oxide structures that show remarkable promise,”
Professor Dibakar Datta – New Jersey Institute of Technology (NJIT)
क्या इसे बनाया जा सकता है?
The team validated their AI-generated structures using quantum mechanical simulations and stability tests, confirming that the materials could indeed be synthesized experimentally and hold great potential for real-world applications.
“These materials have large, open channels ideal for moving these bulky multivalent ions quickly and safely, a critical breakthrough for next-generation batteries.”
Professor Dibakar Datta – New Jersey Institute of Technology (NJIT)
The next step will be to collaborate with experimental labs to synthesize and test the newly conceptualized AI-designed materials.
It could make multivalent batteries the next step in battery technology. So far, the lack of a good material to accept the larger atoms has blocked the development of this option. By using better materials to store magnesium, aluminum, or other large ions, the multi-electron transport capacity of these atoms could maybe one day outshine even lithium’s powerful, but single-electron power-carrying capacity.
सामग्री विज्ञान और AI नवाचार में निवेश
Meta: AI‑संचालित सामग्री विज्ञान
Today, Meta is still mostly known for its Facebook and Instagram social media platforms, as well as WhatsApp chat. It is also present in the virtual reality (VR) space with its VR Headsets and its somewhat unsuccessful “Metaverse”.
Importantly, though, Meta is an AI company with massive investments in infrastructure to make it happen.
“The first multi-gigawatt data center, dubbed Prometheus, is expected to come online in 2026, while another, called Hyperion, will be able to scale up to 5 gigawatts over the coming years.
“We’re building multiple more titan clusters as well. Just one of these covers a significant part of the footprint of Manhattan.
LLM technology seems at first glance to be mostly useful for “talking” tasks, like chatbots, improved online search, education, and other human-centric activities.
(META )
But this research illustrates that the ability of LLM to learn language can be deployed to other data-heavy tasks, like learning how to “talk” crystal structures. The same can be said for genetic codes, for example.
This means that progress in LLM algorithms will likely usher in a golden age of entirely new discoveries in the creation of new materials for batteries, advanced materials, energy generation, etc., as well as new types of proteins and DNA/RNA materials that can be turned into medicine, biomanufacturing tools, etc.
In that context, this implies that companies like Meta and its Lama LLM models are not only building potentially profitable tech replacements for existing tools, but also could turn into an IP powerhouse in the physical world as well.
In that context, it might be remembered that the original tech business of companies like Meta, or for that matter Google (GOOGL ) or Microsoft (MSFT ), was just a stepping stone before turning them into AI & IP-driven giants changing the world with many new technologies, including in renewable energy and material sciences.
नवीनतम Meta (META) स्टॉक समाचार और विकास
संदर्भित अध्ययन
1. Joy Datta, Amruth Nadimpally, Nikhil Koratkar, Dibakar Datta. Generative AI for discovering porous oxide materials for next-generation energy storage. Cell Reports Physical Science, Volume 6, Issue 7, 102665. 16 जुलाई 2025. https://www.cell.com/cell-reports-physical-science/fulltext/S2666-3864(25)00264-4
