Reporting by XinZhiyuan
EDIT: LRS
Google's quantum computing team's technology has always been at the forefront of the world. Recently, they released a summary of their work in 2021, publishing a total of 18 papers, and the small goal of "building a practical quantum computer in ten years" that was once proposed is also steadily implemented.
Quantum computing has always been considered the engine of the next generation of industrial revolution, and various countries and technology companies have added full power to quantum computing and related software for research and development.
Some time ago, Google's quantum AI team also summarized its own 2021.
The researchers say that while quantum computing remains a challenging subject, they have published a total of 18 papers in the past year, yielding fruitful results, especially in building fully error-corrected quantum computers.
And the Quantum AI team has begun work on the next hardware milestone: an error-corrected quantum bit prototype.
Hardware innovation
To do quantum computing, a quantum computer is essential, so the development of hardware is also a top priority.
In May, Google officially announced the Quantum AI Campus in Santa Barbara, which houses data centers, chip manufacturing facilities, research labs and sprawling office areas, all to build and run quantum computer services.
Erik Lucero, the chief engineer of Google's quantum artificial intelligence team at the time, set himself a small goal: to build a error-correcting quantum computer within the next decade.
They also want to use the knowledge and experience gained in developing hardware to develop transformational quantum computer applications.
Google proposes to deliver quantum computers in 2029, mainly because Google really has something in quantum hardware development, which can be roughly divided into the following three points:
1. Google has proven that quantum computers perform better than today's classical supercomputers in specific tasks.
On October 23, 2019, Google claimed to run the mission in about 200 seconds on their quantum chip Sycamore, when summit, the strongest supercomputer on the surface, took 10,000 years to complete the corresponding task. The results were published in the top journal Nature at the time. For a time, the global uproar was widely regarded by the industry as a pioneering milestone in the development of quantum computing, and even compared to the Wright brothers' first aviation flight in Kitty Hawk.
But the term Quantum supremacy is controversial, because Google's claim of 10,000 years actually has moisture, and later researchers gradually reduced the running time of the task to 5 days, and there is no imaginary crushing advantage, so the industry currently uses more quantum superiority to express.
2. Google has the ability to build a prototype of an error-corrected qubit.
Quantum computers, like classical computers, are prone to errors caused by the "noise" of the underlying physical system. How to deal with these errors is a daunting challenge. A normal computer can prevent errors by simply copying bits and using those copies to verify the correct state. But quantum computers can't do this because quantum mechanics forbids copying the unknown state of one qubit to other qubits.
Google physicist Julian Kelly studied the quantum error correction capabilities of the quantum processor Sycamore, which contains a two-dimensional array of 54 superconducting qubits.
The researchers ran two types of quantum error correction codes, one consisting of up to 21 qubits of one-dimensional chain repeat code to test error suppression capabilities, and the other is a two-dimensional surface code composed of 7 qubits as a proof-of-principle experiment for compatibility with larger code settings.
Studies have shown that by increasing the number of qubits based on repeat codes from 5 to 21, the inhibition of logic errors has achieved an exponential increase of up to 100 times. This error suppression ability was stable in 50 error correction experiments.
But Kelly said that despite this, the team was only on their way to completely correcting the error. They failed to solve both of the errors that affect qubits: bit flipping and phase flipping.
Google's current goal is to achieve quantum error correction by redundantly encoding quantum information on multiple physical qubits, proving that this redundancy will lead to improvements using a single physical qubit, which is also the direction of Google's current efforts.
3. Google has the ability to build a logical qubit that has been error-free for any long time.
Logic qubits redundantly encode information across multiple physical qubits and are able to reduce the impact of noise on overall quantum computing. Putting thousands of logic qubits together will enable Google to realize the full potential of quantum computers in a variety of applications.
Error correction qubit progress
For now, the gap between the wide variety of quantum computers and the fully error-corrected quantum computers of the future is still huge.
In 2021, Google has been working on building a prototype logical qubit and reducing its error to an error smaller than the physical qubits on Google's chips, a big step forward for the development of quantum computers.
Getting this done would require improving the entire quantum computing stack, so Google built a chip with better qubits, improved the way it packaged the chips to better connect them to Google's control electronics, and developed the technology to calibrate large chips with dozens of qubits simultaneously.
These improvements resulted in two key outcomes.
First, Google is now able to reset Google's qubits with high fidelity, allowing Google to reuse qubits in quantum computing.
Second, Google implemented mid-circuit measurements, which allowed Google to track calculations within quantum circuits.
In a recent demonstration by Google using repetition codes and exponential suppression of bit and phase flip errors, by using both high-fidelity resets and intermediate circuit measurements, errors were reduced by a factor of 100 as the amount of code grew from 5 qubits to 21 qubits.
Repetitive code is a commonly used error correction tool that enables quantum computers to trade off resources (more qubits) and performance (lower errors), which is also the core guiding ideology of Google to develop future hardware.
In 2021, Google also looked at how errors decreased as the number of qubits contained in 1-dimensional code increased. Google is currently experimenting with extending these results to 2-dimensional surface codes that can correct errors more comprehensively.
Applications of quantum computing
In addition to building quantum hardware, Google's team is also looking for application scenarios where quantum computing has obvious advantages in the real world.
Google has worked with practitioners in academia and industry to explore areas where quantum computers can provide significant computational acceleration, and the expected results are also relevant: error-correcting quantum computers should achieve at least quadratic speedups to be meaningful improvements.
The results of a joint google study with the California Institute of Technology show that under certain conditions, quantum computers can understand physical systems with far fewer experiments than traditionally required. The newly proposed approach, experimentally validated using 40 qubits and 1300 quantum operations, demonstrates huge quantum advantages even with Google's current noisy quantum operations, paving the way for work on quantum machine learning and quantum sensing.
Google, in collaboration with researchers at Columbia University, combined the most powerful chemical simulation technique, Quantum Monte Carlo, with quantum computing, has successfully surpassed previous methods and is now a promising quantum method for ground state many-electron calculations, essential for creating new materials and understanding the chemistry of materials.
Even with noise present on computing devices with up to 16 qubits, Google was able to double the size of previous calculations without sacrificing measurement accuracy.
Google also continues to study how quantum computers can be used to simulate quantum physical phenomena. On November 30 last year, Google published an article in the journal Nature saying that they had created time crystals from Sycamore quantum computing hardware.
This is a great moment for theoretical physicists, the possibility of the existence of time crystals, which they have been thinking about for almost a century.
In other work, Google also worked with collaborators at NASA's Ames Research Center to explore quantum chaotic correlations by experimentally measuring disordered correlations on a quantum computer at Google.
Collaborators with the Technical University of Munich used shallow quantum circuits to create their eigenstates, experimentally measuring the entanglement entropy of the ground state of the Toric code Hamiltonian.
Google said that many of the most influential research results in 2021 were completed with collaborators in various research institutions, and some of them also inspired Google's next research direction.
In 2022, Google Quantum AI will continue to work with other collaborators to explore and implement meaningful quantum applications, quantum chemistry, and multibody quantum physics.
Resources:
https://blog.google/technology/research/2021-year-review-google-quantum-ai/