Given that quantum computing supremacy has been demonstrated on lab, in this article I will review some quantum computing qubit technologies and qubit technologies top challenges.
 |
| Image from Pixabay. |
Quantum supremacy seems to have been demonstrated empirically recently.
On Google Cloud servers, we estimate that performing the same task for m = 20 [where m is the amount of random quantum gates, using 53 qubits] with 0.1% fidelity using the SFA algorithm would cost 50 trillion core-hours and consume 1 petawatt hour of energy.
To put this in perspective, it took 600 seconds to sample the circuit on the quantum processor 3 million times, where sampling time is limited by control hardware communications; in fact, the net quantum processor time is only about 30 seconds.
Are they making a quantum computer?
Progress is mostly being made to make a quantum accelerator, not a full quantum computer.
This accelerator will handle and execute assembly code just like a FPGA or a GPU co-processors, making integration with existing computing languages and their compilers relatively easy, maybe by extending assembly language standard to support this new type of co-processor.
 |
| A CPU would use a quantum accelerator just like a GPU or a FPGA co-processors. |
All the layers above the CPU will behave as any other classical component. There may be a specific classical compiler that will compile code for a quantum accelerator. Above that there may be specific quantum computing programming languages and algorithms, but they will be operated from a classical computer.
The challenges for quantum computing
Some top challenges might be:
- Reliability (Are the quantum circuits doing what we want to do?)
- Most notably, noise and error correction. Noise can cause a qubit to decohere, converting a qubit in a classical bit.
- Error correction does not imply decoherence, but a qubit can still change its state, despite being entangled.
- Scalability
- Can these technologies scale arbitrarily? What are the limits of the technology in this regard?
- Feasibility
- Can we make the manufacture, deployment and maintenance of such systems economically feasible?
Taking those challenges in mind, i'll discuss about some technologies and how they can contribute in overcoming those challenges.
Qubit technologies
Spin qubits (quantum dots)
The spin of a confined electron is used to perform entanglements with other electrons, in a transistor-like circuit, with several transistors serially connected. Electron confinement is done by cooling down up to a point that it is more costly in energy therms for the electron to enter or leave the dot than staying there, and controlling those transistors.Qubits are controlled with magnetic fields.
- Good: Using crossbar technology to connect several arrays of quantum dot, it is believed that it is possible to integrate 1024 qubits in a 30micrometers chip.
- Mild: coherence times are predicted to be in the order of 10^-6 s, and it is thought to be capable of perform about ~10^3 operations. This may be an issue for long-running algorithms.
- Bad: temperatures must be very low, in the order of several mK, in order to avoid noise. This seriously affects feasibility, since it requires costly cryogenic facilities.
NV Center qubits
Nitrogen vacancy center qubits uses a diamond lattice, where a carbon atom is replaced with a nitrogen atom. The qubit consists in the spin of an eletron from the nitrogen atom.
- Good: Coherence times are very long, in the order of seconds.
- Good: May provide additional qubits by interacting with nuclear spins, which may help scale it better, or even error correct the qubit.
- Good: May allow to entangle with farther qubits via photon emissions, even linking them using optics.
- Mild: It can operate at relatively high temperatures, in the order of at least 4K. Still not room temperatures, but much better than mK magnitudes.
Transmon: an example of superconducting qubits
Superconducting qubits consists on multi-level systems where the dynamics are confined to two quantum levels. These devices can be designed as printed circuits. They are not 'physical' qubits like a spin qubit, but more like a virtual qubit.
Circuits consists on islands interconnected with Josephson junctions.
 |
| Josephson junction |
A Transmon qubits a 'charge' qubit
 |
| 2-qubit Transmon processor. DiCarlo et al., 2009 |
These qubits are connected each other by a "Surface code", which consists on a 2 dimensional array of quibits interacting only with the nearest neighbors.
 |
| An array of superconducting qubits. Eleanor Rieffel 2019. |
- Good: quantum supremacy has been proved using this type of qubit.
- Good: In theory they can escalate in size, in a fashion similar to transistors on a CPU.
- Mild: Still requires a superconductor, which typically requires cooling well below 0°C, although it can be interfaced to room-temperature hardware.
- Mild: Operations between non adjacent qubits may require additional gates, thus reducing effective operations on a given decoherence period.
Topological qubits - majorana bound states
A fermion is a type of particle from the standard model whose spin is of the type (x+1)/2, such as quarks.
A majorana fermions a fermion that it is its own anti-particle. It is not known if they exist at all.
However, they could exist as a quasi-state in condensed matter, where for instance the presence or absence of an electron is considered a majorana fermion. This is named a majorana bound state.
Two majorana bound states form a topological qubit.
In order to find majorana bound states, we look for 0 energy states on superconductors.
A majorana fermions a fermion that it is its own anti-particle. It is not known if they exist at all.
However, they could exist as a quasi-state in condensed matter, where for instance the presence or absence of an electron is considered a majorana fermion. This is named a majorana bound state.
Two majorana bound states form a topological qubit.
In order to find majorana bound states, we look for 0 energy states on superconductors.
- Good: Theoretically very long coherence times.
- Mild: Current implementation possibilities require a superconducting medium.
- Bad: Requires near zero energies in order to create the majorana bound states, which means very low temperatures.
Conclusion & final note
Most of these technologies for making a qubit seem promising, perhaps the most unlikely one is the Topological qubit. I am eager to see what does this turn into.
So far a full quantum computer is out of the question, but certainly a quantum accelerator for specific algorithms is viable.
Perhaps the true challenge is finding and implementing quantum algorithms that do require such processing power.
Read more
- QuTech blog
- Quantum Supremacy paper, Rieffel
- Materials in superconducting quantum bits, Olier & Welander.
- Majorana qubits
- Quantiki
Anyway, quantum supremacy should be treated with some skepticism, see https://www.ibm.com/blogs/research/2019/10/on-quantum-supremacy/
ReplyDelete