In a ground-breaking advancement that could redefine the future of data storage, researchers have unveiled a novel laser writing system capable of encoding vast amounts of information within the three-dimensional matrix of glass. This innovative technology achieves unparalleled data density, speed, and energy efficiency, promising long-term archival solutions far beyond the capabilities of conventional storage media. At the heart of this system lies a sophisticated method for creating and controlling minuscule, microscopic voxels—three-dimensional pixels—using ultrafast laser pulses, which encode information directly inside a transparent medium.
The system’s architecture hinges on a femtosecond laser source operating at a 516 nm wavelength, capable of emitting pulses as short as 300 femtoseconds. These ultrashort pulses are meticulously modulated in polarization and amplitude and then scanned across a glass platter measuring 120 mm square and 2 mm thick. Through precision control of the laser’s repetition rate, scanning optics, and sample motion, the research team achieves voxel pitches on the order of submicrons in the lateral dimensions and microns in depth, permitting the stacking of hundreds of layers of information-bearing voxels. This fine-scale modulation and layering culminate in a data density that rivals and exceeds many emerging storage technologies.
Central to the writing process is the manipulation of voxel quality, denoted as Q, which measures bits stored per voxel and is affected by redundant bits added to ensure error correction. The writing system employs two primary voxel encoding regimes: birefringent voxels shaped through a pseudo-single-pulse method, and phase voxels that involve isotropic refractive index changes. The former utilizes a unique combination of beam splitting and polarization modulation, while the latter employs amplitude modulation via acousto-optic modulators to finely tune laser pulses during inscription. Precision in achieving the proper energy per voxel and consistent emission during writing is ensured through feedback mechanisms based on photoemission captured in real time—a closed-loop control system that dynamically balances laser power for stable voxel formation.
A standout feature of this glass storage system is its multi-beam writing capability. By splitting the laser into four modulated beams converging on the objective lens at marginally different angles, the system can simultaneously write in parallel tracks within the glass, significantly enhancing write throughput. This is further optimized by employing polygon scanners spinning at speeds up to 50,000 revolutions per minute and synchronized XYZ translation stages to maintain exact positioning. Together, these components coordinate to deliver peak bit rates in the realm of gigabits per second, a crucial factor for archival data writes where speed and reliability are paramount.
Reading the stored information is accomplished through custom-designed microscopes tailored to each voxel encoding type. For birefringent voxels, a wide-field polarization microscope captures multiple polarization states using liquid crystal variable retarders, maximizing angular information on the polarization state of light passing through the modified glass. This approach reduces overlap and cross-talk between adjacent voxels, enhancing the decoding fidelity. Conversely, phase voxels are read via Zernike phase contrast microscopy, which qualitatively accentuates the refractive index modulations but contends with reduced optical sectioning ability. The researchers mitigate axial cross-talk by acquiring multiple images at differing depths, utilizing contrast oscillations to enhance symbol discrimination.
A crucial aspect of the system’s robustness is the integration of machine learning for symbol decoding. Convolutional neural networks (CNNs) analyze complex image stacks from the microscopy reads, incorporating spatial context and positional encoding to enhance error resilience. This deep learning model processes voxel images, translating raw visual data into probability distributions over symbol states, which in turn inform advanced error correction algorithms using LDPC codes. These codes, derived from the cutting-edge standards in wireless communication, adapt dynamically to varying error rates across sectors on the glass platter, ensuring maximal information retrieval even when some bits are corrupted or lost during reading.
To further push the limits of storage efficiency, the research explores advanced symbol encoding schemes that extend beyond traditional binary Gray codes. By carefully selecting symbol counts and mapping bits over groups of voxels, they maximize channel capacity while minimizing symbol error rates. Optimization techniques identify the ideal modulation levels and the number of discrete symbol states, balancing the competing demands of high bit rate, error correction overhead, and signal-to-noise ratios. This nuanced approach leverages detailed knowledge of the optical channel’s noise characteristics and the physical constraints of the glass storage medium.
Longevity is another hallmark of this technology. The team conducted accelerated annealing experiments at elevated temperatures to simulate the storage medium’s endurance over time. They employed optical diffraction efficiency measurements to monitor the gradual thermal erasure of phase voxels’ refractive index modulations. These macroscopic assessments, combined with quantitative Arrhenius analyses, revealed an activation energy barrier surpassing 3 eV, strongly suggesting remarkable long-term data stability. Such findings highlight the medium’s suitability for archival applications spanning decades, if not centuries, without degradation that plagues magnetic or solid-state devices.
The physical and optical controls entwined in the system—from the fine-tuned lens assemblies correcting spherical aberrations during focus shifts to the polygon scanners ensuring uniform laser pulse delivery—demonstrate a synergy of mechanical precision and optical engineering. Moreover, the feedback algorithms that combat facet reflectivity variations in polygon scanners and environmental fluctuations during writing solidify the system’s reliability. Embedded fiducial markers inscribed into the glass during writing facilitate automated, accurate calibration during reading, enabling seamless location of data sectors and improved read consistency.
With throughputs and densities orders of magnitude beyond contemporary optical storage mediums, this laser-written glass system heralds a new age in archival technology. The integration of ultrafast optics, rigorous physical modeling, lateral and depth-wise voxel control, adaptive decoding, and rigorous error correction converges to provide a scalable, robust, and efficient data repository. As digital archives swell exponentially with global data production, such methods, housed in inert glass that is insensitive to electromagnetic interference, poised to endure geological timescales, could become the foundation of future information preservation strategies.
While this system currently excels at write-once archival tasks, the modular design invites ongoing improvements, such as parallelized multi-beam writing expansions and enhanced neural network architectures for faster, more accurate decoding. Furthermore, its adaptability across multiple glass types and custom optical configurations provides scope for tailored solutions catering to various archival durations, data access speeds, and environmental conditions. These developments point towards a versatile storage medium, far surpassing tapes or hard disks in reliability and ecological sustainability.
In summary, the successful demonstration of ultrafast nonlinear laser writing within glass for massive, dense data encoding spells a transformative step for archival storage technologies. By weaving together innovations in laser physics, materials science, computational imaging, and information theory, this work reveals a pathway toward near-permanent, high-performance data banks. As digitization permeates every facet of modern society, the development of such pioneering technologies will become increasingly vital to safeguarding humanity’s informational heritage against loss and decay.
Subject of Research: Laser-based multi-level data encoding in glass for high-density, durable archival storage.
Article Title: Laser writing in glass for dense, fast and efficient archival data storage.
Article References:
Microsoft Research Project Silica Team. Laser writing in glass for dense, fast and efficient archival data storage. Nature 650, 606–612 (2026). https://doi.org/10.1038/s41586-025-10042-w
Image Credits: AI Generated
DOI: 19 February 2026
Keywords: laser writing, glass storage, archival data, ultrafast lasers, voxel encoding, birefringence, phase contrast microscopy, convolutional neural networks, error correction, LDPC codes, data density, long-term data preservation
