Periodic Reporting for period 2 - CatQubit (Building universal quantum computers with self-correcting cat-qubit technology)
Reporting period: 2023-01-01 to 2024-06-30
Quantum computing is poised to become one of the most transformative technologies of the 21st century. These machines could help us discover new molecules and materials as well as optimize supply chains and investment portfolios. However, current quantum hardware cannot execute computations reliably. Quantum bits (qubits) interactions with the environment introduce errors that compound and throw off computations. To deal with these errors, we can store several copies of each qubit. The computer can then check if there has been an error and correct it. Once this qubit array corrects errors efficiently enough, it becomes what is known as a “logical qubit”. The problem is that to build a single logical qubit, one would need around 1000 physical qubits.
At Alice & Bob (A&B), we discovered how to fast-track quantum error correction, thereby accelerating the development of impactful quantum computers. With our technology, we have the ambition to create an error-corrected quantum computer that will be available in the cloud in just 2-3 years, a decade ahead of our competitors.
Correcting quantum errors is difficult because qubits are subject to two types of errors: bit flips and phase flips. Our novel Schrödinger cat qubits correct bit-flip errors automatically. Having to deal with just one type of error dramatically reduces the number of physical qubits needed for error correction. A logical qubit can be built using fewer than 30 cat qubits, while the competition would require 900-1000 physical qubits.
At Alice & Bob (A&B), we discovered how to fast-track quantum error correction, thereby accelerating the development of impactful quantum computers. With our technology, we have the ambition to create an error-corrected quantum computer that will be available in the cloud in just 2-3 years, a decade ahead of our competitors.
Correcting quantum errors is difficult because qubits are subject to two types of errors: bit flips and phase flips. Our novel Schrödinger cat qubits correct bit-flip errors automatically. Having to deal with just one type of error dramatically reduces the number of physical qubits needed for error correction. A logical qubit can be built using fewer than 30 cat qubits, while the competition would require 900-1000 physical qubits.
In 2022, all the R&D projects of Alice&Bob were focussed on the logical qubit demonstrator with cat-qubits. Since the bit-flip error is corrected at the single cat-qubit level, a logical qubit consists of a 1D chain of cat-qubit that corrects phase-flips. The minimal logical qubit demonstrator consists of 3 data and 2 ancillary cat-qubits. To operate the logical qubit, the ancillary qubits have to be prepared and measured in the X-basis and CNOT gates have to be performed between the data and the ancillary qubits. As for any error correction scheme, these operations have to be performed with sufficient accuracy otherwise, running the repetition code produces more errors than it corrects. When the amount of errors coming from both native errors and imperfect operation of the device equals the amount of errors corrected the device is said to be operating at threshold. As soon as a device is below threshold, error correction is useful and the remaining errors can be scaled down exponentially by increasing the length of the cat-qubit chain. Our 2022 roadmap was divided in 3 main experimental projects :
-Scaling-up the cat-qubit count on a single chip
-Preparing and measuring in the X-basis with high fidelity
-Running the error correction sequence on a multi cat-qubit chip
First, we redesigned the cat qubit to make it more compact to fit on a fixed size chip. This was done by fully redesigning the microwave layout of the cat. The compactness also enabled to increase the non-linearity of the device and its stabilization rate, a critical ingredient for fast gates. As the qubit count increases, our design and microwave simulation tools had to scale-up. Second, the cryogenic set-up was also scaled-up in particular the sample holder shielding the chip and the number of microwave lines available in A&B’s cryostats. For next generation chips, we ramped up our cryogenic power by purchasing several cryostats. Finally, the room-temperature microwave control set-up was up-graded to enable the control of several cat-qubits.
The second project required implementing a set of state of the art techniques and to choose relevant parameter regimes that allow fast and accurate control of the microwave resonator that hosts the cat-qubit. These control techniques involve a intermediate transmon qubit on the chip. We reached state of the art fidelity in the control and readout of this intermediate qubit heavily used in the field. This allowed us to attain a high level of fidelity in the preparation and measurement of cat-states, mostly limited by the finite lifetime of the hosting microwave resonator and finite coherence time of the intermediate transmon qubit.
The third project had two distinct milestones for 2022. The first one was to demonstrate that cat-qubit bit-flip suppression was efficient on a chip comprising several cat-qubits. The second milestone was to work on the implementation and improvement of a CNOT gate between two consecutive cat qubits of the 1D chain. Embedded in the repetition code, this gate maps phase-flips error of the data cat-qubit to the ancillary (the state of the ancillary system is then analysed by the tools developed during the second project).
We also worked on improving single cat-qubits performance for faster gates, using a detuned Kerr approach. This strategy builds on the Kerr cat paradigm where the bit-flip protection comes from Hamiltonian confinement but solves a leakage issue (which eventually leads to bit-flips). A second project designed a parameter regime aimed at improving the repetition code performances. By having some asymmetry between the data and the ancilla cat-qubits one can gain significantly on the error correction threshold.
Finally, we developed our relationship with key distributors of quantum computing power. With Microsoft, for example, we are aligning our efforts towards a shared goal : increased awareness on fault tolerant quantum computation. As a result:
-Microsoft is publishing a tool called "resource estimator" that allows to quantify how qubit noise characteristic impact computation feasibility. It is obviously of great value to our positioning. We were in the private beta group for the tool and published an article on its usage.
-We are now officially member of the Azure Quantum Network.
-Scaling-up the cat-qubit count on a single chip
-Preparing and measuring in the X-basis with high fidelity
-Running the error correction sequence on a multi cat-qubit chip
First, we redesigned the cat qubit to make it more compact to fit on a fixed size chip. This was done by fully redesigning the microwave layout of the cat. The compactness also enabled to increase the non-linearity of the device and its stabilization rate, a critical ingredient for fast gates. As the qubit count increases, our design and microwave simulation tools had to scale-up. Second, the cryogenic set-up was also scaled-up in particular the sample holder shielding the chip and the number of microwave lines available in A&B’s cryostats. For next generation chips, we ramped up our cryogenic power by purchasing several cryostats. Finally, the room-temperature microwave control set-up was up-graded to enable the control of several cat-qubits.
The second project required implementing a set of state of the art techniques and to choose relevant parameter regimes that allow fast and accurate control of the microwave resonator that hosts the cat-qubit. These control techniques involve a intermediate transmon qubit on the chip. We reached state of the art fidelity in the control and readout of this intermediate qubit heavily used in the field. This allowed us to attain a high level of fidelity in the preparation and measurement of cat-states, mostly limited by the finite lifetime of the hosting microwave resonator and finite coherence time of the intermediate transmon qubit.
The third project had two distinct milestones for 2022. The first one was to demonstrate that cat-qubit bit-flip suppression was efficient on a chip comprising several cat-qubits. The second milestone was to work on the implementation and improvement of a CNOT gate between two consecutive cat qubits of the 1D chain. Embedded in the repetition code, this gate maps phase-flips error of the data cat-qubit to the ancillary (the state of the ancillary system is then analysed by the tools developed during the second project).
We also worked on improving single cat-qubits performance for faster gates, using a detuned Kerr approach. This strategy builds on the Kerr cat paradigm where the bit-flip protection comes from Hamiltonian confinement but solves a leakage issue (which eventually leads to bit-flips). A second project designed a parameter regime aimed at improving the repetition code performances. By having some asymmetry between the data and the ancilla cat-qubits one can gain significantly on the error correction threshold.
Finally, we developed our relationship with key distributors of quantum computing power. With Microsoft, for example, we are aligning our efforts towards a shared goal : increased awareness on fault tolerant quantum computation. As a result:
-Microsoft is publishing a tool called "resource estimator" that allows to quantify how qubit noise characteristic impact computation feasibility. It is obviously of great value to our positioning. We were in the private beta group for the tool and published an article on its usage.
-We are now officially member of the Azure Quantum Network.
These three projects have led to several important results for Alice&Bob’s roadmap. The first one has shown our ability to perform a first scale-up of our qubit count on chip as well as of our control equipment. In particular, to our knowledge, our multi-cat qubit chip is the first chip worldwide that comprises several bosonic qubits (cat-qubits are one member of a more general family of qubits stored in harmonic oscillators). Bosonic qubits are more and more praised by the field for hardware efficient error correction and Alice&Bob’s is leading this route, staying focussed on quantum error correction. The second project unlocked state-of-the art control and measurement of cat states, though further work is required to reach below threshold operation. The third project has demonstrated the bit-flip error correction of cat-qubits on a multi cat-qubit chip.
The work done in 2022 led to filing several key patents.
The work done in 2022 led to filing several key patents.