By Andre Saraiva, UNSW on August 16, 2021
A team of researchers from UNSW-Sydney, in Australia, have been developing strategies to control many qubits simultaneously using the same microwave field generated globally, applied to all qubits simultaneously. This development has many layers, described in a series of five manuscripts – the chip itself is modified to have a dielectric cavity laid on top of it, which is simply an engineered piece of crystal (potassium tantalate) that makes the microwaves bounce around and create a homogeneous magnetic field at its bottom. This magnetic field acts on electron spin qubits in a silicon MOS quantum dot system. So, the first layer is a profound engineering revamp of the processor design, which could retire the stripline antennas and re-claim significant on-chip real estate. The researchers have shown that the resonator works as expected, and even went so far as driving two qubits coherently with this global field.
This may sound confusing at first – if the field is irradiated on all qubits at the same time, how could one control each qubit separately to perform process information and implement algorithms? This is the second layer of this innovation step – the qubits are now to be individually addressed by controlling their natural frequency separately. While spins all have a natural frequency set by a homogeneous external DC magnetic field, small variations (less than 1%) occur naturally by an effect called spin-orbit coupling. This effect is described by Einstein’s relativity theory, which states that the same electric fields created by the gate electrodes used in confining the quantum dots can also be used to control the electron spins. This means that no additional control lines are required. Therefore, qubit can be tuned in and out of resonance with the global field from the cavity. Variations of this type of local addressing under a global field were conceived even as early as the invention of silicon quantum computers themselves, by Kane in 1998. The UNSW team, however, is rethinking exactly how this local control can be done.
The second layer of this development is precisely on the control strategy, which employs qubits that are always driven (instead of removing the qubits completely from resonance as originally envisioned by Kane and others). One could compare the challenges of operating many qubits in a quantum processor with the circus art of plate spinning. Qubits left on their own will eventually loose coherence, and one trick to prolong its life is to give it a little swirl, which in the jargon is called dynamic decoupling. This trick is well known, and periodic dynamic decoupling steps are incorporated in most quantum processors. The UNSW team, however, figured out how to continuously drive the qubits, a trick known as “dressing the qubits”, without ruining all other operations in a quantum processor. Single isolated dressed qubits had been studied before, but this is the first time that this trick is harmonised with all tasks necessary for universal quantum computing.
The third layer is the killer. Reliance on spin-orbit coupling is difficult because materials properties in solid state devices vary from qubit to qubit. Noise is everywhere. Quantum founding father Wolfgang Pauli even went as far as saying “One shouldn’t work on semiconductors, that is a filthy mess” – a quote that obviously did not age well. The UNSW team developed, however, a method to dress qubits in a modulated field and control qubits by synchronously shifting their frequencies. This protocol is referred by the acronym SMART (Sinusoidally Modulated, Always Rotating and Tailored).
The technology that works well for a few bits might not be ideal for a large processor. Those readers who were using computers in the 1980’s will appreciate the revolution that CMOS computers represented – while individual NMOS and PMOS transistors worked fine, once a large number of transistors were integrated the necessity to have homogeneity, tolerance to noise and dissipate less heat became so important that transistors had to be reimagined from scratch. This is the type of thinking that currently keeps quantum engineers busy – how do we move from fussy few qubits to sturdy systems that can be collectively controlled in a simplified manner?
The SMART protocol addresses this issue by incorporating noise deterring strategies, as discussed in this theory paper. Indeed, the advantages of this strategy were experimentally demonstrated in a traditional device (without the dielectric crystal at the time), and qubit coherences were pushed from 16 microseconds (on a bare qubit) to 235 microseconds (for the dressed qubit) to a record 2 milliseconds in the case of SMART qubits. This more than hundred-fold improvement in coherence time, however, needs to be understood in the context of qubit control time, which is now much slower. While slower control is desirable for high fidelity classical electronics, it also means that less qubit gates can be performed, so the best metric to be observed is the bottom-line gate fidelity, which was measured to exceed 99%. This confirms that these gates are within the tolerance for quantum error correction.
So, what’s next?
From theory to experiments there is always a gap, and the size of this gap is only partially known. The demonstration of the SMART protocol in a single qubit is insufficient to confirm that the advantages promoted by this technique will really resist the variability in a large number of qubits. Moreover, the dielectric cavity needs to be tested in a chip that is designed specifically for this purpose in order to gain a real understanding of its potential for coherent control of a large number of spins.
Collected at: https://quantumcomputingreport.com/qubits-in-unison-innovations-in-global-control/