The research team’s central methodological innovation is deceptively simple but conceptually powerful. Most prior studies of doped LiCoO2 have reported improvements in lithium-ion mobility or conductivity without separating the contribution of lattice expansion from that of electronic structure modification. When a larger guest atom is inserted into the crystal, it pushes neighboring layers apart, widening the pathways through which lithium ions travel during charging and discharging. That geometric effect alone can speed up ion diffusion, even if the dopant does nothing else. By explicitly subtracting this lattice expansion effect from their calculations, the researchers could isolate the intrinsic electronic consequences of each dopant, providing a cleaner picture of what is really happening inside the cathode at the atomic scale.

Using first-principles calculations, a computational approach that derives material properties directly from quantum mechanics without fitted parameters, the team examined two co-doping schemes in detail. The first pairs sodium and magnesium, both placed at the lithium sites of the layered structure. The second pairs magnesium and aluminum, which substitute primarily at the cobalt sites. This site-selected design was deliberate: because magnesium can occupy either lithium or cobalt sites depending on synthesis conditions, comparing the two schemes allowed the researchers to disentangle what each dopant does in each crystallographic position, a question that has muddied the experimental literature for years.

The sodium-magnesium results delivered the study’s most striking finding. When the lattice expansion contribution was subtracted out, the co-doped material still showed significantly improved electron conductivity, and the lithium-ion diffusion was promoted specifically through the expansion of the spacing between lithium layers. In other words, the two dopants act synergistically: sodium, being a large monovalent ion, acts as a structural pillar that props open the lithium interlayers, giving lithium ions wider corridors to migrate through, while magnesium contributes an electronic boost that raises the material’s electrical conductivity. Neither effect alone would be as effective, and the study demonstrates that the pairing works because each element addresses a different bottleneck in the cathode’s operation.

This division of labor has deep roots in the physics of layered oxides. LiCoO2 adopts a layered structure in which lithium ions, cobalt ions, and oxygen atoms occupy alternating planes. During battery discharge, lithium ions shuttle between the cobalt-oxygen layers, and their mobility depends on the width of the gaps between those layers. Meanwhile, the electrons that flow through the external circuit must move through the cobalt-oxygen framework itself, so the electronic conductivity of that framework controls how fast the cathode can deliver current. A cathode that only solves one of these problems hits a wall at the other, which is precisely why single-element doping strategies have produced such inconsistent results across the literature.

The magnesium-aluminum calculations reinforced this picture from a different angle. Because magnesium can also occupy cobalt sites, the team performed co-doping calculations with magnesium and aluminum at those positions. The results confirmed a clean functional split: magnesium primarily enhances conductivity, while aluminum stabilizes the crystal structure, and both effects operate through influencing the electronic structure rather than through geometric changes. This finding aligns with decades of experimental observations. Magnesium doping of LiCoO2 has long been known to enhance electronic conductivity, a result reported as far back as the late 1990s, and aluminum doping has been repeatedly shown to improve the structural robustness of the cathode, particularly at the high charging voltages where undoped LiCoO2 begins to degrade.

That voltage threshold is where the commercial stakes become vivid. Standard LiCoO2 cathodes are typically charged to around 4.2 volts or slightly above, but extracting more lithium at higher cutoff voltages would substantially increase energy density, which is why researchers worldwide are racing to stabilize LiCoO2 at 4.6 volts and beyond. Above roughly 4.5 volts, the layered structure undergoes destabilizing phase transitions, cobalt dissolution accelerates, and oxygen vacancies can form during delithiation, leading to microcracking and rapid capacity fade. The new study adds an important variable to this race: calculations of the average intercalation potential indicate that magnesium-aluminum co-doping is superior to sodium-magnesium co-doping in optimizing the cutoff voltage, making the cobalt-site strategy the more promising route for high-voltage applications.

Perhaps the most practically consequential conclusion, however, concerns doping levels. The calculations imply that a low-content co-doping strategy is a better solution for LiCoO2-based cathode design. This recommendation runs counter to a naive intuition that if some doping is good, more must be better. Excessive dopant concentrations can disrupt the delicate electronic band structure that makes LiCoO2 such an effective cathode in the first place, and heavy substitution can introduce structural defects that outweigh the benefits. The study’s subtraction methodology strengthens this conclusion, because it shows that the beneficial electronic effects of the dopants persist even after the purely geometric contributions are removed, meaning designers can achieve the desired improvements with minimal disruption to the host lattice.

The work, which was supported by the National Key Research and Development Program of China, arrives at a moment when LiCoO2 research is experiencing a renaissance driven by demand for high-energy-density batteries in smartphones, laptops, drones, and electric aircraft. Cobalt-rich layered cathodes remain unmatched in volumetric energy density and tap density, and recent advances, including magnesium-pillared LiCoO2 stable at 4.6 volts, gradient aluminum doping with nanofilm coatings enabling cycling at 4.7 volts, and ternary inert-element co-doping approaches, have pushed the material’s limits dramatically. What the Huazhong team contributes is a methodological discipline: a way to attribute performance gains to their true physical origins, so that future doping recipes can be designed rationally rather than discovered by trial and error.

For battery engineers, the takeaway is a clearer map of the design space. If the goal is faster charging and higher rate capability, opening the lithium interlayers with pillar-like sodium while magnesium handles the electronic conductivity offers a synergistic lithium-site strategy. If the goal is pushing the charging voltage higher without structural collapse, magnesium and aluminum at the cobalt sites provide conductivity and stability in tandem while optimizing the intercalation potential. And in both cases, restraint is a virtue: low dopant concentrations capture the electronic benefits while preserving the intrinsic qualities that have kept LiCoO2 at the heart of lithium-ion technology for more than three decades. By teaching the field to subtract the obvious geometric effect and look at what remains, the study turns a long-standing ambiguity into an actionable design principle for the next generation of high-voltage cathodes.

Subject of Research: First-principles analysis of site-selected Na, Mg, and Al co-doping effects on lithium-ion diffusion, electronic conductivity, and structural stability in LiCoO2 cathodes

Article Title: Site-selected Na, Mg and Al co-doping effects in cathode material LiCoO2 by subtracting the contribution of lattice expansion

Article References: Zhang, H., Hou, C., Tang, S., Cao, Y., Wang, Z., & Ouyang, Z. (2026). Site-selected Na, Mg and Al co-doping effects in cathode material LiCoO2 by subtracting the contribution of lattice expansion. Ionics. https://doi.org/10.1007/s11581-026-07502-2

Image Credits: AI Generated

DOI: 10.1007/s11581-026-07502-2

Keywords: LiCoO2, cathode materials, co-doping, first-principles calculations, lithium-ion batteries, electronic structure, lattice expansion, sodium doping, magnesium doping, aluminum doping, high-voltage cathodes, lithium-ion diffusion

The software, described in the journal SoftwareX, is the work of Yasuaki Ito and colleagues, who built it as a lightweight companion to their GPU-native research code GANSU. Where GANSU targets CUDA-equipped supercomputers and research-sized molecules, GANSU Lite targets the hardware a person already owns: a laptop, a tablet, or even a smartphone. The distinction matters because, as the authors note in their survey of the existing landscape, no prior package combines a full self-consistent-field (SCF) engine, density-functional theory, post-Hartree–Fock correlation methods, analytic derivatives, Hessians, and vibrational analysis while running entirely on the client side. Web front ends such as WebMO and CalcUS hide the installation behind institutional servers; browser viewers such as Jmol and 3Dmol.js render molecules but solve nothing. GANSU Lite is, by its authors’ account, the first tool to do both.

Two features of the modern web platform make this possible. The first is WebAssembly, a binary instruction format that lets developers compile code written in fast systems languages—in this case Rust—and run it in the browser at near-native speed. GANSU Lite’s arithmetic-heavy kernels, including the electron-repulsion integrals, Fock matrix builds, molecular-orbital transformations, and coupled-cluster amplitudes that dominate the wall time of a calculation, are compiled Rust modules shipped as static assets. The second is the Web Worker, a genuine background thread that keeps the interface responsive. The entire SCF driver runs off the main UI thread, so a multi-second calculation never freezes the page; a live panel reports the current iteration, and users can cancel a running job at any time without reloading.

The architecture splits cleanly into two layers. A TypeScript layer handles everything the browser sees: the document structure, the Three.js scene graph that renders and rotates molecules, session storage, and the SCF control flow with its convergence tests and DIIS bookkeeping. The Rust layer owns the dense floating-point work—the two-electron integrals that describe how every pair of electrons interacts, the effective potentials assembled from them, and the derivative machinery for gradients and Hessians. Within each SCF iteration, one-electron integrals are computed in TypeScript while the two-electron block dispatches among three classical integral-evaluation schemes—McMurchie–Davidson, Obara–Saika, and Rys quadrature—chosen by the angular momentum of each shell quartet, with Schwarz screening and eight-fold permutation symmetry trimming the workload. Resolution-of-identity approximations, which factorize the two-electron integrals to cut their cost, are available for Coulomb, exchange, and MP2 terms.

The deployed site exposes eight task-focused pages. A free-form single-point calculator accepts molecules from a built-in library of more than two dozen species—ranging from hydrogen and small ions up to benzene, naphthalene, and vitamin C—or from a pasted or uploaded XYZ file. Seven curated pages each isolate one concept: basis-set convergence plots total energy against basis-set label; a Walsh diagram page sweeps bond angles and overlays orbital-energy curves; a vibrational page computes harmonic frequencies and simulated infrared spectra; a potential-energy-scan page stretches bonds; a multi-method page places Hartree–Fock, MP2, and CCSD side by side; and pages for geometry optimization and atomic charges round out the set. Seven standard basis sets, from the minimal STO-3G to the triple-zeta def2-TZVP, are selectable with a row of buttons.

The method list reads like a syllabus for computational chemistry. Restricted, unrestricted, and restricted-open-shell Hartree–Fock; Kohn–Sham DFT with LDA, GGA, meta-GGA, and hybrid functionals including B3LYP and PBE0; Grimme’s D2 dispersion correction; MP2 and MP3 perturbation theory; coupled-cluster singles and doubles; and three excited-state methods—CIS, ADC(2), and TDDFT—that predict the wavelengths at which molecules absorb light. Every calculation runs to convergence in the browser; nothing is precomputed on a server. Results arrive as tables of orbital energies, Mulliken and Löwdin charges, and Wiberg bond orders, alongside plots drawn on HTML canvas: convergence curves, potential-energy profiles, and simulated spectra.

Three-dimensional orbital isosurfaces, too expensive to render comfortably in the calculator itself, are delegated to a sibling tool called MOrbVis. With one click, GANSU Lite serializes the converged wavefunction as a standard Molden record, writes it into the browser’s sessionStorage, and opens MOrbVis in a new tab, which reads the record and renders interactive orbitals with positive and negative lobes and cross-section cuts. Because the handoff travels through sessionStorage rather than a network, the wavefunction never leaves the browser and is discarded when the tab closes—preserving the platform’s zero-upload guarantee. And because Molden is a standard interchange format, the same wavefunction can be opened in desktop tools such as Molden or Avogadro.

The benchmark results are striking. On four devices—a Windows desktop, a Windows laptop, an Apple-silicon laptop, and an iPhone 16 Pro Max—the heaviest test workload, a Hessian calculation for carbon dioxide, completed in 17 to 36 seconds, while single-point and optimization tasks finished in under five seconds everywhere. At a fixed 150-basis-function workload, the phone actually outran both Windows machines, completing in 91.9 seconds against 152.5 and 175.4 seconds, because the kernel is single-threaded WebAssembly with SIMD instructions, so a fast single core beats many slower ones. Correctness was verified against the established PySCF package: Hartree–Fock and MP2 total energies for small molecules in three basis sets agreed to within two millionths of a Hartree. WebAssembly’s pinning of floating-point arithmetic to a well-specified IEEE-754 subset means the same calculation yields essentially identical answers across browsers and operating systems—a quiet but consequential win for reproducibility.

The limits are honest and well mapped. The two-electron integrals are held in memory, and their storage grows as the fourth power of system size, so calculations hit WebAssembly’s 32-bit address-space ceiling of 4 gigabytes at roughly 200 basis functions—regardless of how much RAM the machine has. The iPhone fails earlier and for a different reason: iOS reclaims tabs that hold too much memory for too long, illustrating that on mobile, being able to allocate memory and being able to sustain it are different things. The single-threaded design is itself a deliberate trade-off, since multi-threaded WebAssembly requires cross-origin isolation headers that are awkward to provide on plain static hosts. The authors are candid that GANSU Lite trades peak throughput for reach; it does not replace research-grade packages, and they do not claim it does.

What the platform removes instead is the barrier between curiosity and computation. A lecturer can compute and display a result live during a talk on whatever machine the room provides. A reader can check a published number by opening a URL rather than installing a package to reproduce a single data point. An author can attach a live, re-runnable demonstration to a paper, letting anyone inspect a claim on any device. In classrooms and outreach settings, where provisioning software on every machine is impractical or forbidden, every participant can run the same calculation from a shared link. The enabling idea is simple and, in retrospect, obvious: the calculation travels as a URL, and the computation happens on whatever hardware is already at hand. Released under the BSD-3-Clause license with a public demo and repository, GANSU Lite turns the browser tab—arguably the most universally available computing environment ever deployed—into a working quantum chemistry laboratory.

Subject of Research: A zero-install, browser-based quantum chemistry platform built on WebAssembly and Web Workers

Article Title: GANSU lite: A zero-install, browser-based quantum chemistry platform

Article References: Ito, Y., Fujii, H., Tsuji, S., Nakano, K., & Kasagi, A. (2026). GANSU lite: A zero-install, browser-based quantum chemistry platform. SoftwareX, 36, Article 103046. https://doi.org/10.1016/j.softx.2026.103046

Image Credits: AI Generated

DOI: 10.1016/j.softx.2026.103046

Keywords: quantum chemistry, GANSU Lite, WebAssembly, electronic structure, Hartree–Fock, density-functional theory, coupled cluster, browser computing, vibrational analysis, Hiroshima University, open-source software, computational chemistry