ConnexionLaser-written glass plates could store data for thousands of years
Humans are generating more data than ever before. While much of these data do not need to be stored long-term, some – such as scientific and historical records – would ideally still be retrievable in decades, or even centuries. The problem is that modern digital archive systems such as hard disk drives do not last that long. This means that data must regularly be transferred to new media, which is costly and time-consuming.
A team at Microsoft Research now claims to have found a solution. By using ultrashort, intense laser pulses to “write” data units called phase voxels into glass chips, the team says it has created a medium that could store 4.8 terabytes (TB) of data error-free for more than 10::000 years – a span that exceeds the age of history’s oldest surviving written records.
Direct laser writing
The idea of writing data into glass or other durable media with lasers is not new. Direct laser writing, as it is known, involves focusing high-power pulses, usually just femtoseconds (10-15 s) long, on a three-dimensional region within a medium. This modifies the medium’s optical properties in that region, and each modified region becomes a data-storage unit known as a voxel, which is the 3D equivalent of a pixel.
Because the laser’s energy is focused on a very small volume, the voxels created with this method can be very densely packed. Changing the amplitude and polarization of the laser’s output changes what information gets encoded at each voxel, and an optical microscope can “read out” this information by picking up changes in the light as it passes through each modified region. In terms of the media used, glass is particularly promising because it is thermally and chemically stable and is robust to moisture and electromagnetic interference.
Direct laser writing does have some limitations, however. In particular, encoding information generally requires multiple laser pulses per voxel, restricting the technique’s throughput and efficiency.
Two types of voxel, one laser pulse
Microsoft Research’s “Project Silica” team says it overcame this problem by encoding information in two types of voxel: phase voxels and birefringent voxels. Both types involve modifying the refractive index of the medium, and thus the speed of light within it. The difference is that whereas phase voxels create an isotropic change in the refractive index, birefringent voxels create an anisotropic change by rotating the voxel in the plane of the 120-mm square, 2-mm-thick glass chip.
Crucially, both types of voxel can be produced using a single laser pulse. According to Project Silica team leader Richard Black, this makes the modified region smaller and more uniform, minimizing effects such as light scattering that can interfere with read-outs from neighbouring voxels. It also allows many voxel layers to be written into, and then read out from, a single glass chip. The result is a system that can generate up to 10 million voxels per second, which equates to 25.6 million bits of data per second (Mbit s−1).
Performance of different types of glass
The Microsoft researchers studied two types of glass, both of which have better mechanical properties than ordinary window glass. In 301 layers of fused silica glass, they achieved a data density of 1.59 Gbit mm−3 using birefringent voxels, with a write throughput of 25.6 Mbit s−1 and a write efficiency of 10.1 nJ per bit. In 258 layers of borosilicate glass, the data density reached 0.678 Gbit mm−3 using phase voxels. Here, the write throughput was 18.4 Mbit s−1 and the write efficiency 8.85 nJ per bit.
“The phase voxel discovery in particular is quite notable because it lets us store data in ordinary borosilicate glass, rather than pure fused silica; do it with a single laser pulse per voxel; and do it highly parallel in close proximity,” says Black. “That combination of cheaper material and much simpler and faster writing and reading was a genuinely exciting moment for us.”
The researchers also showed that they could directly inscribe the glass using four independent laser beams in parallel, further increasing the write speeds for both types of glass.
Surviving “benign neglect”
To determine how long these inscribed glass plates could store data, the team repeatedly heated them to 500 °C, simulating their long-term ageing at lower temperatures. The results of these experiments suggest that encoded data could be retrieved after 10::000 years of storage at 290 °C. However, Black acknowledges that this figure does not account for external effects such as mechanical stress or chemical corrosion that could degrade the glass and the data it stores. Another unaddressed challenge is that storage capacity and writing speed will both need to grow before the technology can compete with today’s data centres.
If these deficiencies can be remedied, Black thinks the clearest potential applications would be in national libraries and other facilities that store scientific data and cultural records. “It’s also compelling for cloud archives where data is written once and kept indefinitely,” Black says. He points out that the team has already demonstrated proofs of concept with Warner Bros., the Global Music Vault and the Golden Record 2.0 project, a “cultural time capsule” inspired by the literal golden records launched on the Voyager spacecraft in the 1970s.
A common factor across all these organizations, Black explains, is that they need media that can survive “benign neglect” – something he says Project Silica delivers. He adds that the project also provides what he calls operational proportionality, meaning that its costs are primarily a function of the operations performed on the data, not the length of time the data are kept. “This completely alters the way we think about keeping archival material,” he says. “Once you have paid to keep the data, there is little point in deleting it, and you might as well keep it.”
Microsoft began exploring direct laser data storage in glass nearly a decade ago thanks to team member Ant Rowstron, who recognized the potential of work being done by physicist Peter Kazansky and colleagues at the University of Southampton, UK. The latest version of the technique, which is detailed in Nature, grew out of that collaboration, and Black says its capabilities are limited only by the power and speed of the femtosecond laser being used. “We have now concluded our research study and are sharing our results so that others may build on our work,” he says.
The post Laser-written glass plates could store data for thousands of years appeared first on Physics World.
