Data, DNA and diamonds

We live in a data age. So much so that the numbers describing our data consumption are now too large to grasp. For example, 2.5 quintillion bytes of data are created every day (that’s a one with 18 zeroes after it). This rate is only increasing, with 90 per cent of all the data that currently exists being generated in just the last two years.

Research group the International Data Corporation (IDC) predicts we will be consuming 163 zettabytes of data a year by 2025 (a zettabyte is a trillion gigabytes). To put that into perspective, the current total amount of data on the entire internet is estimated at just under three zettabytes.

All that information must be stored somewhere, and that takes space and money – two resources that are starting to feel the pinch. “Storing data is increasing at a rate faster than the cost is coming down by technological improvement,” says Dr Nick Goldman of the European Bioinformatics Institute. “The people, like you and me, who aren’t paying just haven’t felt the pain yet.”

Goldman is well placed to understand the increasing financial pressure of data storage as he has experienced it first-hand. The European Bioinformatics Institute stores data about DNA and genomes generated by scientists all around the world to the tune of 120 petabytes (a petabyte is a million gigabytes).

Concerns about how to store that data and the increasing cost of doing so led Goldman and his colleagues to think about alternative methods. They settled on a substance that was close to hand and available in vast quantities – DNA. “Each of us have a machine that weighs something in the region of 50-100kg that produces vast quantities of DNA,” says Goldman. “The input is essentially a few thousand calories of energy and a bit of light and that can make vast amounts.”

After a few back-of-an-envelope calculations, Goldman and his colleagues figured DNA could, in principle, be competitive in terms of price, stability and storage density (they calculated that all the data in the world, if stored in DNA, would fit in the back of a van) and that it was achievable with current methods.

They set about encoding a series of media on strands of DNA, including a photograph of the European Bioinformatics Institute, a 26-second MP3 fragment of Martin Luther King’s famous ‘I Have A Dream’ speech, and a PDF of Watson and Crick’s paper describing the structure of DNA, all totalling 1MB.

The process they used is disarmingly simple. The 0s and 1s in which digital data is stored are converted into the four chemical bases that make up DNA – A (adenine), C (cytosine), T (thymine) and G (guanine). “DNA, the hard disk drive of living organisms, is fundamentally a digital system,” says Goldman. “There are four different types of DNA base and they will join together in a highly stable chain in any order, and in principle these can be exceedingly long chains of information millions of sub-units long.”

The code is then sent to a commercial biological engineering company which synthesises DNA strands that match the sequence of digital code. To read the information back, the DNA is simply run through a sequencer to retrieve the code, which can be translated back into binary.

Today the process of building DNA strands can’t match the millions-long chains found in nature, instead producing individual strands of just a few hundred nucleotides. In practice, this means that data stored on DNA needs to be cut up and put back together like a jigsaw puzzle.

Each strand of DNA also requires a tiny piece of indexing data which specifies where it fits into the jigsaw. This reduces the amount of actual data each strand can carry, but fortunately not enough to restrict relatively large amounts of data storage.

“If you want to store twice as much information, it’s only a constant additional amount of indexing,” says Goldman. “So, it does grow, but better than linearly. In reality, that means we’re fine for quite a long time.”

The main bottleneck is the DNA synthesising process, which is costly and time-consuming. Five years ago, it cost Goldman and his colleagues $25,000 to synthesise enough DNA to encode 1MB of information. Now he says the cost is down by a factor of ten – still far too expensive to be commercially viable but falling at a promising rate.

“In the year 2000 it cost $3bn to sequence a human genome,” Goldman says. “Now it’s down to $1,000. So, we’ve dropped six orders of magnitude in 10 years and that’s going on with no sign of stopping.”

Still, Goldman doesn’t see DNA replacing hard drives or memory sticks. Rather he views it as a form of long-term archival storage for the most valuable data like bank records, national historic archives or information about the position of nuclear waste. Eventually it might trickle down to the person on the street, backing up valuable data such as family photos and videos.

DNA is the ideal medium for this because of its longevity. All it requires is a cool, dry, dark environment and it can survive undamaged for thousands of years, as evidenced by DNA taken intact from the remains of 20,000-year-old mammoths.

‘DNA, the hard disk drive of living organisms, is fundamentally a digital system.’

DNA isn’t the only new or exotic system for storing data long-term. Chemical methods have allowed researchers from the University of Ghent in Belgium to store information in the form of powder.

The team found a way to transfer binary information into the chemical signature of a sequence-defined macromolecule.

The researchers wrote an algorithm to translate the 0s and 1s of binary into a 15-letter language. They then cut up the string of data and added indexing information on how to put it back together, much like the DNA technique. This sequence of data was then chemically synthesised and could be read back by a technique called tandem mass spectrometry. The group wrote another algorithm to speed up the reading process and finally were able to successfully encode a QR code in powder form.

With the QR code previously only being 1089 bytes in size, this process is a relative newcomer to the field and a long way behind DNA in terms of research. However, Steven Martens, a researcher from the Ghent team, believes the process could compete with DNA in terms of longevity and storage density. “We can keep it stable for years without a problem,” says Martens. “We believe we can make simpler structures but also longer. In DNA you play with four units. We play with 15 units, so you would need a larger DNA structure to write the same thing we did.”

Martens sees the method as another long-term archival application, but chemically stored data could have other, more imaginative, applications. A team under Jean-Francois Lutz at the Institut Charles Sadron in Strasbourg is looking at ways of using it to combat counterfeiting.

The method involves encoding barcodes into polymers, which are then inserted into the fabric of a product to be trademarked, like a running shoe. “This barcode in the material proves it is correct,” says Martens. “So, if I bought shoes, I can prove in the very structure of my shoe that it’s really Nike.”

The future of long-term data storage looks promising, but what about the more readily accessible data that we need on a daily basis – our hard drives, disks and memory sticks? “The holy grail in memory technology is to have a universal memory,” says Professor Geoffrey Beach of the Massachusetts Institute of Technology (MIT).  What’s needed, he explains, is “a kind of memory that is very high density – something like a hard disk drive – but something very high speed, something that is not mechanical. If you had some kind of material whose physics are allowed to achieve all of these properties, there could be tremendous benefits.”

Beach and his team think they have found such a material, or materials, and it all revolves around a tiny particle that’s not really a particle at all.

Skyrmions are newly discovered ‘quasi-particles’ of magnetism. They’re called quasi-particles because they’re not real particles at all, but the cumulative effects of electrons’ spins oriented within a magnetic material. These effects produce particle-​like characteristics such as size and mass. Beach describes them as twists in the magnetisation of a material in a way that is analogous to crop circles.

“If you imagine a field of grass and all the grass blades are pointing up,” he explains, “if you twist some of these around, that will look like a particle. In fact, you could take that twist and move it around without the grass itself actually moving.”

What makes skyrmions especially exciting is that they are twists of magnetism that, once twisted, cannot be undone, making them inherently stable. These quasi-particles can then be manipulated and controlled using electric fields.

Another property that makes skyrmions so promising is that they can be harnessed in everyday materials like iron and platinum. “These are materials that we’ve known about for centuries now,” says Beach. “They’re some of the simplest materials you can imagine. But when you put these materials together and interface between them you get entirely new properties.”

This means exotic electromagnetic effects like skyrmions can be produced without the extreme low-temperature conditions that semi­conducting materials normally require.

Beach’s research is at the cutting edge of an existing field called spintronics, which manipulates the spin of electrons to switch magnetic bits on and off. There are functioning spintronic devices already on the market in which the power consumption is almost zero. Beach hopes that future skyrmion-based devices will benefit from the same low-power processes.

Combining this with the other benefits of skyrmion-based devices could provide the holy grail that researchers have been searching for. “What we would have is a device typically with as high as or even significantly higher data density than a magnetic hard drive,” he says, “but with no mechanical operation, with very high speed and with low power consumption. That would really quite likely be a default data storage device.” In other words, the hard disk drive of the future.

The potential doesn’t stop there. Because the control of electron spin and the flow of magnetism is so great in skyrmion-based systems, it could be harnessed to form logic devices as well, and that could lead to nothing less than a revolution in computing.

“Currently you have transistors that do all the thinking of a computer,” explains Beach. “You have memory devices that store that information and they have to talk to each other. One of the bottlenecks is piping that data back and forth. If you had some material or device that could process the information and store it within the same device, now you’ve taken the memory and the logic, and you’ve put that into one architecture.”

That doesn’t mean there aren’t hurdles to overcome. The issue of reading data encoded with skyrmions has not yet been solved in an efficient commercialisable way. However, Beach believes the read-out problem is not insurmountable, pointing out that similar solid-state magnetic memory devices have already solved the problem of reading data. It just needs to be modified to skyrmion-based devices – an engineering problem rather than a fundamental physical one.

Beach believes it could be 10 years before a skyrmion-based memory device is on the market, but other applications such as neuromorphic computing (mimicking the brain’s computational hardware) could be a commercial reality even earlier. Whether it’s sooner or later, Beach thinks the technology could ultimately be a game changer.

“One can continue processing and storing data more densely and lower cost and faster and continue on a line,” Beach says. “But the step changes that happen in history are when there’s a new technology that’s qualitatively different, not just a little better, and I think that’s what we’re trying to look at.”