IBM Makes Breakthrough in Race to Commercialize Quantum Computers

[F-22Raptor]

Researchers at International Business Machines Corp. have developed a new approach for simulating molecules on a quantum computer.

The breakthrough, outlined in a research paper to be published in the scientific journal Nature Thursday, uses a technique that could eventually allow quantum computers to solve difficult problems in chemistry and electro-magnetism that cannot be solved by even the most powerful supercomputers today.


In the experiments described in the paper, IBM researchers used a quantum computer to derive the lowest energy state of a molecule of beryllium hydride. Knowing the energy state of a molecule is a key to understanding chemical reactions.

In the case of beryllium hydride, a supercomputer can solve this problem, but the standard techniques for doing so cannot be used for large molecules because the number of variables exceeds the computational power of even these machines.

The IBM researchers created a new algorithm specifically designed to take advantage of the capabilities of a quantum computer that has the potential to run similar calculations for much larger molecules, the company said.

The problem with existing quantum computers – including the one IBM used for this research -- is that they produce errors and as the size of the molecule being analyzed grows, the calculation strays further and further from chemical accuracy. The inaccuracy in IBM’s experiments varied between 2 and 4 percent, Jerry Chow, the manager of experimental quantum computing for IBM, said in an interview.

Alan Aspuru-Guzik, a professor of chemistry at Harvard University who was not part of the IBM research, said that the Nature paper is an important step. “The IBM team carried out an impressive series of experiments that holds the record as the largest molecule ever simulated on a quantum computer,” he said.

But Aspuru-Guzik said that quantum computers would be of limited value until their calculation errors can be corrected. “When quantum computers are able to carry out chemical simulations in a numerically exact way, most likely when we have error correction in place and a large number of logical qubits, the field will be disrupted,” he said in a statement. He said applying quantum computers in this way could lead to the discovery of new pharmaceuticals or organic materials.

IBM has been pushing to commercialize quantum computers and recently began allowing anyone to experiment with running calculations on a 16-qubit quantum computer it has built to demonstrate the technology.

In a classical computer, information is stored using binary units, or bits. A bit is either a 0 or 1. A quantum computer instead takes advantage of quantum mechanical properties to process information using quantum bits, or qubits. A qubit can be both a 0 or 1 at the same time, or any range of numbers between 0 and 1. Also, in a classical computer, each logic gate functions independently. In a quantum computer, the qubits affect one another. This allows a quantum computer, in theory, to process information far more efficiently than a classical computer.

The machine IBM used for the Nature paper consisted of seven quibits created from supercooled superconducting materials. In the experiment, six of these quibits were used to map the energy states of the six electrons in the beryllium hydride molecule. Rather than providing a single, precise and accurate answer, as a classical computer does, a quantum computer must run a calculation hundreds of times, with an average used to arrive at a final answer.

Chow said his team is currently working to improve the speed of its quantum computer with the aim of reducing the time it takes to run each calculation from seconds to microseconds. He said they were also working on ways to reduce its error rate.

IBM is not the only company working on quantum computing. Alphabet Inc.’s Google is working toward creating a 50 qubit quantum computer. The company has pledged to use this machine to solve a previously unsolvable calculation from chemistry or electro-magnetism by the end of the year. Also competing to commercialize quantum computing is Rigetti Computing, a startup in Berkeley, California, which is building its own machine, and Microsoft Corp. which is working with an unproven quantum computing architecture that is, in theory, inherently error-free. D-Wave Systems Inc., a Canadian company, is currently the only company to sell quantum computers, although its machines can only be used to solve certain optimization problems.

https://www.bloomberg.com/news/arti...gh-in-race-to-commercialize-quantum-computers

 

[F-22Raptor]

Inside the clean room at Rigetti Computing's Fab-1 facility in Fremont, California.


First quantum computers need smart software


The world is about to have its first quantum computers. The complexity and power of quantum hardware, such as ion traps and superconducting qubits, are scaling up. Investment is flooding in: from governments, through the billion-dollar European Quantum Technology Flagship Program, for example; from companies, including Google, IBM, Intel and Microsoft; and from venture-capital firms, which have funded start-ups. One such is ours, Rigetti Computing, which in June opened the first dedicated facility for making quantum integrated circuits: Fab-1 in Fremont, California. The vision is that commercial quantum- computing services will one day solve problems that used to be unimaginably hard, in areas from molecular design and machine learning to cybersecurity and logistics.1

The problem is how best to program these devices. The stakes are high — get this wrong and we will have experiments that nobody can use instead of technology that can change the world.

We outline three developments that are needed over the next five years to ensure that the first quantum computers can be programmed to perform useful tasks. First, developers must think in terms of 'hybrid' approaches that combine classical and quantum processors. For example, at Rigetti we have developed an interface called Quil2, which includes a set of basic instructions for managing quantum gates and classical processors and for reading and writing to and from shared memory. Second, researchers and engineers must build and use open-source software for quantum-computing applications. Third, scientists need to establish a quantum-programming community to nurture an ecosystem of software. This community must be interdisciplinary, inclusive and focused on applications.

Hybrid systems
Today's quantum programming differs from much previous theoretical work on algorithms; it is becoming more and more practical.

Theoretical computer scientists have been developing potential algorithms for imagined quantum computers since the 1990s. Mathematician Peter Shor's famous code for breaking encryptions was one of the first; many more are listed in the Quantum Algorithm Zoo from the US National Institute of Standards and Technology (see go.nature.com/2inmtco). These algorithms are generally designed for big, noiseless quantum computers, which are unlike the devices that will be available within the next five years. These will have tens to thousands, not millions, of qubits, with little redundancy to correct for internal errors. They will calculate a limited range of things in a noisy way. For example, they will not be able to use Shor's algorithm to find the prime factors of large numbers. So their use must be targeted: they will not always beat conventional computers.

These limitations can be overcome by building quantum processors as 'accelerators' to boost the performance of conventional computers. A classical computer might, for example, optimize operations to compensate for noise in the quantum processor, or aggregate answers from sequences of short quantum programs. Such hybrid programming has been demonstrated in quantum chemistry3 and in optimization4. Algorithms that run on small, superconducting quantum processors have performed steps in calculating the ground states of materials and molecular systems, for example5, 6. Another algorithm has solved constrained optimization problems, which are common in areas such as machine learning, logistics and scheduling4.

We've found, however, that it can be hard to predict the performance of hybrid algorithms. For example, the quality of the quantum subroutine in hybrid algorithms for chemistry can vary greatly depending on the system that is being simulated and the mathematical tricks used. So hybrid quantum-computing algorithms need to be studied empirically, as they are for machine learning. The way to find out how a system works is to build it, see what it does and back up any rules of thumb with mathematics later. This work will begin in earnest once the first quantum computers are available, and it will accelerate fast.

To reach this stage, researchers must change their mindsets, and this could be hard. We will find that some past work has little utility. We've all seen talks on quantum algorithms whose complexities are peppered with huge exponents, meaning that they could take millions of years to complete. For the coming devices, such codes are so impractical as to be useless.

Quantum programmers must care about practical details such as noise models and exact counts of logic gates. They will have to decide which qubits in the computer to use and how to deal with ranges of operational fidelities and low-level precisions that are foreign to most modern programmers. But the gain will be worth the pain.

In turn, hardware designers need to be responsive to the choices and preferences of quantum programmers, so that their technology can become more useful.

Open software
Different classical computers behave similarly enough to enable software written for one to run on others. Early quantum computers will have their own nuances, and software for them will need to be bespoke. When each operation and instruction matters, generalized solutions need to be optimized, and software and hardware designed concurrently. Algorithms must be discovered numerically rather than algebraically, and developed using simulators and software rather than pens and paper.

Innovative digital tools are needed for developing and testing algorithms, writing software and programming the devices. Quantum programmers should keep an eye on the underlying physics, so that they are aware of different types of noise in sequences of pulses, for example. Performance benchmarks, such as a suite of standard molecules to simulate, are also necessary.

Differences between quantum and classical programming begin at the instruction level. Classical computers use Boolean logic — with basic operations such as AND, NOT, OR. Operations in quantum computations, such as multiplying tensors and matrices, are much more complex and result in unusual behaviour. For example, quantum information cannot be cloned exactly between processor registers; and reading the state of a quantum register alters the information stored in it.7 Hybrid software needs to handle all these behaviours simply enough for programmers to be able to code easily. The result will be a new programming paradigm, as object-oriented, probabilistic and distributed programming once were.

Quantum programmers must decide which aspects of the system are essential for them to consider and which they can skim over in practice. For example, executing a program on superconducting quantum processors requires instructions to be translated many times. Control and readout instructions are converted from digital to analog to quantum to analog to digital as they go from the control hardware to the qubits and back. Programmers don't want to have to deal with all the microwave engineering and physics, but they need to be aware of how these processes affect noise or the time it takes to run the code. They need tools to work directly with the devices, so that they can understand and exploit the trade-offs.

Easy programming interfaces are crucial to making quantum computers widely usable; examples include Quil and OpenQASM8 from IBM. More sophisticated options still need to be added, such as optimizations for specific types of processors. Higher-level languages for writing and compiling quantum programs also need to be developed.

It is important that all these tools are open source. Such a model was not available at the dawn of digital computing, but its power to speed innovation, as with Linux in the early days of the web, is essential for the quantum-programming community to grow quickly. We have made a start with our quantum-programming toolkit, Forest, which is written in Python, open source and accessible to anyone. It joins an exciting early ecosystem — much of it open source — developed by different academic and industrial research groups. Other examples are LIQUi|> (embedded in F#), Scaffold (C++), Quipper (Haskell), QGL (Python), ProjectQ (Python), QCL, QuIDDPro and Chisel-Q (Scala). Researchers must resist pressure to standardize tools prematurely or keep the high-level, exploratory parts of the programming stack proprietary.

Build a community
A new breed of quantum programmer is needed to study and implement quantum software — with a skillset between that of a quantum information theorist and a software engineer. Such programmers will understand how quantum devices operate well enough to instruct them and minimize problems. They will be able to build usable software and will have a deep knowledge of the mathematics of quantum algorithms and computation. Experts from fields in which the software will be applied must be closely involved if the code is to be truly useful. For example, chemists such as Alán Aspuru-Guzik at Harvard University in Cambridge, Massachusetts, drove interest in using hybrid algorithms in quantum-chemistry calculations. Researchers in other fields, especially in machine learning and optimization, should get on board.

Advanced kinds of education are needed to train this new breed. Several centres are well positioned to draw together the interdisciplinary skills and tools needed to offer degrees in quantum-computer engineering: the Institute for Quantum Computing at the University of Waterloo in Canada, the Institute for Quantum Information and Matter at the California Institute of Technology in Pasadena, the quantum-engineering doctoral training centres in the United Kingdom, and QuSoft, the Dutch research centre for quantum software in Amsterdam. At Rigetti we have started a Junior Quantum Engineer programme for bachelor's degree students, which includes training in quantum programming. We have partnered with the Quantum Machine Learning accelerator at the Creative Destruction Lab (a technology-transfer centre that fosters start-ups) at the University of Toronto, Canada, to provide access to and support for Forest and other programming tools.

Early-career quantum programmers have tremendous opportunities to become leaders of a transformational field. But they need support. Their supervisors must recognize that work on an open-source software project might delay their next pure research paper. They need industrial internships to gain a broader practical perspective. And they need institutional backing to work between the fields of software engineering and quantum physics.

Next steps
It is crucial that research on quantum-computing algorithms is tied more closely to research on the software that's used to implement them.

First, funders should insist that theoretical work is implemented in software and, as much as possible, tested on hardware. Second, algorithm researchers must be explicit about the architecture they are targeting. They must show evidence of how algorithms will be practically implemented on different noisy systems. Third, funders and journal editors must establish standard ways to assess algorithm performance and resource requirements. This will enable hardware and software to improve together, and will sift out the most viable algorithms more quickly. Open-source tools should be used wherever possible, and publications should encourage the publication of code alongside results.

Finally, the quantum-computing community must prioritize engagement with experts in areas such as simulation and machine learning. Quantum and classical programmers must collaborate more. We call on every current and aspiring quantum-algorithm researcher to present their work at a classical conference at least once in the next year. It falls to us to expand the community that will realize the incredible potential of quantum computing.

http://www.nature.com/news/first-quantum-computers-need-smart-software-1.22590

IBM Has Used Its Quantum Computer to Simulate a Molecule—Here’s Why That’s Big News

We just got a little closer to building a computer that can disrupt a large chunk of the chemistry world, and many other fields besides. A team of researchers at IBM have successfully used their quantum comptuer, IBM Q, to precisely simulate the molecular structure of beryllium hydride (BeH2). It's the most complex molecule ever given the full quantum simulation treatment.

Molecular simulation is all about finding a compound's ground state—its most stable configuration. Sounds easy enough, especially for a little-old three-atom molecule like BeH2. But in order to really know a molecule's ground state, you have to simulate how each electron in each atom will interact with all of the other atoms' nuclei, including the strange quantum effects that occur on such small scales. This is a problem that becomes exponentially harder as the size of the molecule increases.

While today's supercomputers can simulate BeH2 and other simple molecules, they quickly become overwhelmed and chemical modellers—who attempt to come up with new compounds for things like better batteries and live-saving drugs—are forced to approximate how an unkown molecule might behave, then test it in the real world to see if it works as expected.

The promise of quantum computing is to vastly simplify that process by exactly predicting the structure of a new molecule, and how it will interact with other compounds. In work published today in Nature (paywall)—and also avilable on the Arxiv (PDF)—the IBM team have shown that they can use a new algorithm to calculate the ground state of BeH2 on their seven-qubit chip.

In some ways, it's a small advance. But it's an important step on the path of ever-greater complexity in molecular simulation using quantum computers that will ultimately lead to comercially important breakthroughs.

Even now, as the research team notes in their blog post on the work, IBM offers access to a 16-qubit quantum computer as a free cloud service. The more qubits a chip has—that is, quantum bits that can be used to encode data in multiple states at once—the greater the complexity of calculations it should be able to handle. At least in theory. As we pointed out when we made practical quantum computers one of our Breakthrough Technologies of 2017, one of the big challenges in designing quantum computers is making sure qubits remain in their delicate quantum state long enough to perform calculations. The more qubits a chip has, though, the harder that has been for researchers to do.

Still, the day when quantum computers surpass classical machines—an inflection point known as quantum supremacy—is rapidly approaching. Some observers think a chip with 50 qubits would be enough to get there. And while the chemistry world stands to benefit immensely from such advances, it isn't the only field. Quantum computers are expected to be superstars at any kind of optimization problem, which should help propel big advances in everything from artificial intelligence to how companies deliver packages to customers.

https://www.technologyreview.com/th...r-to-simulate-a-molecule-heres-why-thats-big/

 

[Hamartia Antidote]

http://www.sciencealert.com/ibm-has-simulated-the-most-complex-molecule-yet-with-a-quantum-computer

IBM Just Broke The Record of Simulating Chemistry With a Quantum Computer

Engineers have modelled the interactions between subatomic components of a complex molecule using a quantum computer, making a significant leap forward in our modelling of chemical reactions.

The simulations were carried out by IBM on superconducting hardware, and this milestone just pushed into new territory for what can be achieved using quantum computing.

The molecule in question was beryllium hydride – or BeH2. It's not the fanciest molecule in town, but there's still a lot going on between those two hydrogens and single beryllium for a computer to figure out.

Last year, Google engineers simulated the bonding of a pair of hydrogen atoms on its own quantum computer, demonstrating a proof of principle in the complex modelling of the simplest arrangement of energies in molecules.

Molecular simulations aren't revolutionary on their own – classical computers are capable of some pretty detailed models that can involve far more than three atoms.

But even our biggest supercomputers can quickly struggle with the exponential nature of keeping track of quantum interactions of each new electron involved in a molecule's bonds, something which is a walk in the park for a quantum computer.

These revolutionary devices have been big news of late, with big players in the information technology world investing heavily in the race for quantum supremacy – the line in the sand where quantum computers become truly practical tools that surpass the power of traditional computing systems.

For a quick 101; quantum computers are devices that use a particle's binary states in specific kinds of calculations, much like a 1 or 0 in binary code.

Specifically, this property has a blurred in-between state called a superposition, the nature of which can be applied in calculations that would take a classical computer a long time to run through.

This makes quantum computers a big deal for some things, such as finding supersized prime numbers or – as in this case – crunching the numbers on particle interactions within a molecule.

Unlike those solar-system style diagrams your high school chemistry teacher drew on the board, electrons don't behave like little spheres whizzing around a nucleus.

Instead they exist in a mind-bending state of possibilities that only get more complicated as you add more particles into their surroundings.

This constitutes what's called a many-body problem in physics, and even just a few particles in one or two dimensions demands some hardcore problem solving.

Usually physicists will find short-cuts. One such simplification, for example, is called the Monte Carlo method, which applies a statistical sampling process to solve rule-based problems.

When it comes to increasing numbers of charged particles, these kinds of short cuts can quickly fall apart.

Having a working quantum computer can potentially provide a neat way to avoid these problems.

The things is, even the latest quantum computers are small and prone to making mistakes.

As cutting edge as it is, the seven qubit device used in this study still relied on delicate states that could only be used in calculations for microseconds, leaving little time for lengthy processes.

The goal was to come up with an efficient algorithm that would describe the arrangement of particles in molecules with three atoms, including lithium hydride and beryllium hydride.

"Our scheme contrasts from previously studied quantum simulation algorithms, which focus on adapting classical molecular simulation schemes to quantum hardware – and in so doing not effectively taking into account the limited overheads of current realistic quantum devices," IBM researchers explain on their blog.

Ultimately this means we'll be better prepared for the next generation of quantum computers that tackle bigger molecules.

It's hoped that one day, we'll have such detailed solutions to various many-body problems that we'll be able to predict the interactions of compounds far more accurately, pointing the way to improved drugs or spotting obscure side effects before clinical trials even begin.

Eventually, the sky will be the limit for quantum computers. But even the biggest and best devices will be giant paperweights without the right software to drive them.