Enlarge /. Honeywell's ion trap hardware.
Over the years, scientists have developed a variety of systems on which you can run quantum algorithms. Most of them had one or two helpful properties – easy to manipulate or be able to maintain for a longer time – but the others were lacking enough to keep them from being practical computing solutions. In recent years, however, a number of companies have figured out how to make a significant number of solid-state qubits called transmons. Because the manufacturing technology for transmons is similar to that of existing chip manufacturing, many of the key players in the emerging market – including Google, IBM and Rigetti – have opted for transmons.
But transmons aren't ideal either. They require extremely cold temperatures, show significant device-to-device variability, and are good but not good at maintaining their condition. A number of people in the field I've spoken to have suggested that there is room for another technology to outperform transmons, and Ars & # 39; s own Chris Lee is spending his money on it.
Now a company that is new to the quantum computer market is also betting on it. Honeywell, a company better known as a defense company and material supplier, announces that it has built a quantum computer with an alternative technology called "ion trap" and will make it available later this year through Microsoft's Azure cloud service. The company also claims that by some standards it is the most powerful quantum computer to date, but this claim needs to be weighed very carefully.
Transmon qubits circulate a current through a loop of superconducting wire, which is connected to a resonator, with which the current state can be controlled and read out. However, both the superconducting wire and the resonator have to be manufactured, which allows subtle differences between the individual qubits. In addition, all hardware must be kept extremely cold – within a tiny fraction of a percent of absolute zero – to keep these relatively large objects close to their quantum ground state.
Trapped ions provide a way to overcome some of these challenges. The actual qubit consists of a very small number of atoms – in Honeywell's case only two. Tony Uttley, President of Quantum Solutions at Honeywell, stressed to Ars that this eliminates manufacturing problems because each device has identical properties that are defined by the atom used (in this case ytterbium). "Every qubit starts perfectly," said Uttley to Ars, "every mistake you make is a mistake you make because of the surrounding infrastructure."
Enlarge /. According to Honeywell, the know-how in support hardware led to the quantum computer. And there is a lot of support hardware.
Based on Honeywell's experience in building and integrating this infrastructure, the company's engineers feel in an excellent position to minimize this noise. The other thing is that small clusters of atoms like this can be cooled with lasers. While the environment must be kept very cold, it does not have to reach the extreme temperatures required by a Transmon.
In Honeywell's case, the ytterbium ions were not particularly easy to cool with lasers, so they threw a few barium ions into the mixture and cooled them with lasers. The four ion ensemble was easy to cool and control, and the environment only needed to be kept at 12K. This requires liquid helium, but not the complex dilution cooling systems required by Google and IBM hardware.
Because they are charged, the ions can be moved within the device simply by changing the local electric fields using the approximately 200 electrodes located along the device. The state of the electrons of the ion can be manipulated with lasers at certain wavelengths, whereby electrons can be superimposed on potential energy states. Entanglement and various gate operations can be achieved simply by moving two ions nearby and using laser operations that manipulate both at the same time. Reading is achieved by stimulating the ions with another laser, whereby the ions emit a photon that indicates their state.
Honeywell hardware can be viewed as a linear sequence of different devices. Ions enter at one end and are mixed through alternating areas where they can be held for storage or hit by lasers that perform qubit manipulation operations on them. A gate operation (the quantum equivalent of performing an AND or NOT operation) can be performed by simply placing two ions in the same location and performing an operation on both at the same time. In addition, clusters of four ions (two ytterbium, two barium) can be split in half or two clusters of two ions can be brought together.
The device that Honeywell describes today arranges four qubits along a single line of these memory / manipulation levels. However, the diagram of the device also shows two additional lines of storage and manipulation stages flanking the line used in these initial experiments. This is in line with what Uttley Ars said: Honeywell is convinced that the device can be scaled quickly, with the expectation that additional qubits can be added every year without fundamentally changing the architecture. While four qubits are quite low compared to Transmon devices (more on that in a moment), the company believes it can close the gap very quickly.
Enlarge /. There are a lot of lasers involved in controlling the states of these qubits.
An interesting aspect of this setup, which Uttley says is not currently available on any other commercial system, is that you can measure a qubit individually without necessarily interfering with anything else in the system. (Technically, this is done through a science fiction sounding operation called a quantum teleportation CNOT gate.) This allows the computer to execute the equivalent of an "if" instruction and the algorithm based on the results of measuring that single qubit to change. After the measurement, the qubit can also be reset and reused for further calculation.
Individual components of it work very well. A possible problem are so-called "condition preparation and measurement errors", which have adopted the acronym SPAM. In this case, Honeywell researchers found that the SPAM is dominated by measurement errors, which occur in less than 1 percent of cases. Single qubit gates have errors that are an order of magnitude lower, and two qubit gates are at a similar level. All of this is significantly lower than the typical behavior of Transmon-Gates.