Breadcrumb navigation

The top researcher of quantum annealing talks about its importance and the future prospects contributing to environmental issues.

A two part interview with Dr. Gabriel Aeppli, who greatly contributed to the development of the quantum annealing, about its academic appealing and future prospects. In the first part, he talked about his research area, his research team's impact on D-Wave, that commercialized the world's first quantum annealing machines, and the advantages of quantum computers over classical computers. In this second part, he explains quantum annealing itself in an easy-to-understand manner, talks about the differences from simulated annealing as well as the future prospects of quantum computing in general, and gives messages to young researchers.

Quantum annealing is like a shortcut by digging a tunnel in the mountain

Q1Could you briefly explain "quantum annealing"?
A1

Before explaining quantum annealing, let's talk about simulated annealing. It was when I started working at Bell Laboratories that I read the famous paper of Scott Kirkpatrick (and colleagues Vecchi and Gelatt) on simulated annealing.
For example, consider a combinatorial optimization problem, such as the traveling salesman problem. This involves finding the best route for salesmen to meet customers in different towns. Expenses such as transportation costs, wages during the traveling time, and accommodation expenses at hotels are critical for the business which employs the salesman, so it makes sense to find a route with which he can visit all towns, and meet all customers at the minimum cost. But such problems, whose results depend on too many factors, cannot be rapidly solved with conventional algorithms on classical computers.
To formulate their approach, Kirkpatrick and his co-researchers were inspired by natural phenomena such as cooling hot soup and crystallization. Above its boiling point, all ingredients are randomly distributed, but on cooling, the soup approaches it lowest internal energy state, with ingredients having higher density accumulating at the bottom and supernatant gathering on the surface. This fact attracted their attention and they realized that the process of cooling down a complex system such as a soup with a mix of ingredients, is in a way similar to solving a problem like that of the traveling salesman.
So, the idea was born that the combinatorial optimization problem could be solved by the same methods as those used to simulate the cooling of hot systems on a classical computer. This is simulated annealing.
On the other hand, Feynman’s thoughts mentioned earlier in this interview, suggested that the thermal cooling could be replaced by quantum cooling.
For the soup, once heated, many configurations of the ingredients are sampled and slow cooling selects a state which has internal energy lower than neighboring states, which we can hope is not far from the overall minimum energy of the system. In the combinatorial optimization problem, the possible configurations of the system are sampled via quantum fluctuations (*5) instead of thermal fluctuations (*6), and then by removing the quantum fluctuations, the system then naturally settles to an equilibrium state with minimum internal energy. The idea that the system naturally equilibrates and minimizes the energy is the basic concept of quantum annealing.

  • (*5)
    Quantum fluctuation: A term in quantum physics that refers to the temporary change in energy at a certain point in space.
  • (*6)
    Thermal fluctuation: A term of statistical mechanics that refers to random deviations from the mean state in a system at equilibrium. The thermal fluctuations increase as the temperature increases.

Q2Can you explain in more detail the difference between quantum annealing and simulated annealing?
A2

Well, there are steep mountains in both Switzerland and Japan. Consider a terrain with several peaks and valleys. And suppose going from the valley at this end to the lowest valley, somewhere beyond the peaks. That valley represents the minimum energy state.
Simulated annealing, which simulates heating and cooling, is like a hike. The trajectory is to climb up the first peak, then cross it down to the next valley. Then, as you continue to cross the next peak and descend to the valley ahead, you will eventually reach the lowest valley somewhere.
On the other hand, the trajectory of quantum annealing is like passing through tunnels connecting the valleys, so that you can go straight to the lowest valley. This tunneling mechanism is a major feature of quantum annealing.

Q3What was the reason you had expectations for quantum annealing 20 years ago?
A3

That is the knowledge obtained through actual experiments. Quantum annealing based on quantum cooling gave different results than simulated annealing utilizing thermal cooling: it seemed to pick out minima in sharper valleys. This then suggested that for certain combinatorial optimization problems (namely those characterized by energy landscapes with sharp peaks and valleys), quantum annealing would work more efficiently than classical annealing.

Our world will be transformed by quantum computing

Q1Can you tell us about the future possibilities and your prospects of quantum annealing?
A1

A major application of quantum annealing, provided that decoherence times and entanglement lengths (*7) achievable in the hardware are both suitably long, will be in the simulation of quantum physical properties. Since phenomena in the field of chemical reactions and biochemistry are basically derived from quantum physical properties, they can be applied, for example, to find the optimal solution when performing chemical synthesis for purposes such as pharmacology.
On the other hand, some have argued that quantum annealing machines may not offer dramatic improvements in processing speed compared to classical computers, and this debate has not been settled, although these workers also promote the (not inconsistent) idea that implementing quantum annealing on classical computers actually does lead to improved results for certain optimization problems. Using the metaphor of the peaks and valleys like before, what if the mountains were like rolling hills in Germany? For example, the yields of portfolios of certain private investments and government bonds in politically stable countries may be represented by rolling hills. In that case, classic hiking may be easier to get to the destination valley.
Related to that, determining what problems are appropriate for quantum annealing, simulated classical annealing and gate-based quantum computing remains a scientific challenge. Regarding factorization of large numbers, gate-based quantum computing is still considered to be advantageous, but it has not been proved mathematically and research is still ongoing. In the past, proofs of mathematical conjectures have taken more than 200 years, so this may be a challenge of that kind.

  • (*7)
    decoherence times and entanglement lengths: technical terms which measure quantum-ness in time and space.

Q2How will quantum computing change our world?
A2

It will have a significant impact on the concept of data security and on the study of quantum physical properties occurring in chemistry and molecular biology. There will be major changes in the world as chemical and biochemical simulations become more accurate.
Also, as mentioned earlier, if a true quantum computer is built which does not have large classical (electronic and thermal) overheads, the problems of IT related energy consumption and carbon footprint will be solved, and this should have a huge impact on our society as well.

Expectations for young researchers

Q1Quantum computing is a very promising area, so please give some advice for young researchers challenging this area?
A1

My advice is to consider, first of all, that the winning technology has not been selected yet.
For example, in the world of quantum computing, traditional approaches from the semiconductor domain have included two platforms. One is electron spin-based platforms, which I am working on, and the others exploit the Josephson effect in superconducting systems.
In addition, atomic physics approaches using trapped ions or trapped atoms (*8) are popular, but all gate-based approaches, whether rooted in solid state or atomic physics, aiming at realization of general-purpose quantum computers seem to face great difficulties.
Therefore, young researchers will find any platform challenging and interesting!
I also want them to keep energy consumption in mind. Ultimately, quantum computers' greatest contribution to mankind will be to increase the efficiency of computing. That's what the IT industry can do to reduce global warming. This is my key message.

  • (*8)
    trapped ion or trapped atom: A certain type of quantum computer system using atoms that are ionized and cooled to stop their movement and operate with a laser. The ionized atom in that state is called "trapped ion" or "trapped atom".

Q2Some people think that media coverages and industry's expectations for quantum computing are overhyped. What is your view?
A2

I don't think that – on average - they are overhyped anymore, with most current reports on the topic modulating enthusiasm with an appreciation of the long timescales likely required for real success. While coverage has been good, I even hope it should attract more attention, because talent and various kinds of support and funding are needed to actually advance research. I think it's good to be interested in this field, and I'm glad to see people have big expectations for quantum computing.
And finally, our initial papers on quantum annealing were the result of efforts by not only myself but also three other physicists, two of whom were at the University of Chicago and another who moved from UC to NEC during the project. I would like to add that it is collaborations such as this, where different participants bring different strengths, which enable the greatest progress on difficult technical problems.