For many, the theoretical physics question Canela is pursuing is a challenging concept, however the applications or perhaps implications of the research are something with potentially wide-ranging effects.
Victor Canela arrived on New Zealand’s shores three years ago. His sole reason for coming to New Zealand – being able to work on one of physics’ biggest questions – how do we get from A to B, when we don’t even know where B is?
Victor says, "During my undergraduate degree, I learned about quantum mechanics and the strangeness of the microscopic universe and became very interested in quantum optics: the study of the interaction between light and matter in the tiniest of scales – photons and atoms. Coming to New Zealand to work with Prof. Carmichael at the University of Auckland has been a great experience as he has done fundamental work on the field."
"The system we study is called a one-atom maser, similar in operation to a laser, with microwaves instead of light. We simulate a stream of two-level atoms (nicknamed quantum bits or qubits in analogy with the classical computer bits) interacting with photons in a box. We are interested in controlling the properties of the photons by modifying properties of the stream of qubits."
It is not surprising that a young researcher would choose this field. In 2012, Serge Haroche and David Wineland were recipients of the Nobel Prize for Physics for experiments that realised the delicate interaction of matter and light.
Canela himself is very good at explaining the fundamentals of his research:
"The difficulty in studying and simulating these systems is that none of them are closed; they have losses and in general have a complex interaction with the outside world. Computationally speaking, the simulations are too complex to handle. A way around this is to introduce random fluctuations to our system and thus mimic the effect the outside environment would have. These are the so-called quantum trajectories."
"Individual quantum trajectories are created with random processes, so we don't know in advance where they will go. What we do know is that their ensemble average gives us information of properties of the system that we are interested in, such as how many photons are inside the box."
When run in serial, one trajectory after another, it would take about 1200 core-hours to obtain useful data (nearly two months on your standard desktop computer). Canela was already aware of NeSI’s Pan cluster and had optimised his code for running in an HPC system. Upon attending a Hacky Hour hosted by team members of the University of Auckland’s Centre of eResearch and NeSI, Canela met NeSI’s Jordi Blasco. It was during one of these Hacky Hours that Blasco and Canela were able to optimise Canela’s code, speeding up the processing time by a factor of 150.
Blasco explains, "We realised that the problem could easily transform into a highly decoupled problem with parameter switching workflow. In less than one hour working with the Fortran code, the application was able to run several independent trajectories at once."
Thanks to the Intel VTune Amplifier reports, they were able to identify the major bottlenecks in the code. After a few meetings optimising parts of the code, Canela and Blasco finally achieved an overall improvement in speed of up to 600 times faster.
The results so far have shown some interesting insights:
"We have been successful in simulating the interaction between qubits and photons and have devised ways to control the photon number. Such quantum control is of big interest to the community as it provides tools for the development of a quantum computer."
The video below shows three individual sample photon number trajectories, for a mean atom number of 0.65. Also shown is the average of 1000 of these trajectories - note that the average has no sudden jumps, it is a more or less smooth line converging to a definite photon number value. The individual trajectories visit particular photon numbers more often, these are the so-called "trapping states". The trajectories jump between these regions and provided you average enough of them, converge to a mean photon number for a particular mean number of atoms.
Findings such as those above are only the beginning for Canela’s thesis:
"The control mechanisms we have developed have allowed us to simulate the creation of states with particular photon statistics and we have been able to explore different parameters regimes, thanks to the NeSI Pan Cluster. We aim to expand our simulations and include more sources of noise such as one would have in an experiment."
