Silicon enjoys an unrivaled lead in human fabrication know-how and investment. Importantly, in its isotopically purified form, it is an ideal neutral material to host sensitive spin qubits. Proponents emphasise the potential of this technology to deliver chip-scale solutions where some other approaches would require ‘building-size’ machines. But spin quibts are delicate. Nanofabricating the right environment to extend their coherence life, and so improve fidelity, has been a challenge.
Electron spin is the archetypal quantum two-level system. Silicon is an ideal substrate in which to confine electrons. Various approaches have been investigated for this purpose, including quantum dots and donor phosphorus atoms.
A growing range of promising approaches are now reaching liftoff. These vary widely in terms of the fabrication technology they seek to employ, from the CMOS mainstream to ultra-cutting edge atomic level precision manufacture.
For a detailed recent review of silicon spin qubits see .
- 28Si has no nuclear spin, so it forms an ideal magnetically neutral substrate to host sensitive spin qubits. Good spin qubit lifetimes are possible in isotopically pure material.
- Silicon fabrication technology is unrivalled by that of any other material.
- Fast gates have been demonstrated, 0.8-80ns.
- Nanoscale qubit footprints have been realised.
- Multiple specific technologies to confine spin qubits within the silicon substrate are being investigated: Donor phosphorus impurities, MOS Q Dots, Si/SiGe Q Dots and Ge/SiGe Q Dots.
- Cryogenic cooling is required, but some silicon spin variants promise operation in the more moderate 1K regime.
- CMOS fabrication promises enhanced opportunities for integration of the quantum plane, control plane and control processors. Techniques for control multiplexing are starting to emerge .
- Potential to leverage techniques for microwave control and cryogenic systems already being developed for use with superconducting qubits.
- Large scale FTQC machines could have a compact footprint.
- Isotopic purification of natural silicon is required to improve qubit coherence lifetimes. Current experiments typically use material purified to 800ppm, this may need to be further improved .
- 2Q gates have not yet been demonstrated with fidelity above 99%.
- Only very small numbers of qubits operated so far; potential cross-talk problems have not yet been investigated.
Key Players and Approaches
A wide variety of approaches for forming spin qubits in silicons are being developed.
Silicon with MOS Quantum Dot Spin
Conduction electrons are confined in a silicon quantum dot formed at the silicon/silicon-dioxide interface of a Metal-Oxide-Semiconductor (MOS) structure, localised with a metal gate electrode.
Notable Commercial Players: Intel/QuTech, Quantum Motion, Origin Quantum, Hitachi
- Directly leverages CMOS fabrication techniques.
- 1Q fidelities in lab 99.96% .
- 2Q fidelities in lab 98.0% ; anticipated that higher frequency gates will allow 99%+.
- Long qubit state lifetime (T1) up to 0.1s .
- At the relatively high temperature of 1K, 1Q gates with fidelity of 99% have been demonstrated , and a qubit lifetime (T1) of 1ms ; 2Q gates have also been demonstrated at 1K with indications that high fidelity should also be possible .
- New geometries are emerging to mitigate leakage errors .
- Qubits are formed directly at the Si02 interface, a potential source of fabrication defects and noise .
The potential for this technology to leverage CMOS manufacturing techniques and operate at the relatively ‘mild’ temperature of 1K is striking. Coupling between spin qubits and microwave photons is likely to be possible (although demonstrated only in SiGe quantum dots so far). Multiple groups around the world are pursuing this technology, notably Intel in partnership with QuTech. Recent UK startup Quantum Motion is also targeting this technology.
Fact Based Insight was very surprised to see SQC recently announce it was no longer working on this variant of silicon spin qubit technology , despite a recent run of record breaking announcements by Andrew Dzurak . Identification of a CMOS fabrication partner to take this forward is likely to require the facilitation of a major government programme.
Silicon with SiGe Quantum Dot Spin
Conduction electrons are confined in a silicon quantum dot formed at a silicon/silicon-germanium heterostructure interface, localised with one or more metal gate electrodes. The use of an epitaxialy grown and strained silicon provides a very low disorder lattice.
Notable Commercial Players/Consortia: Intel/Qutech
- SiGe has been an increasingly regular adjunct to CMOS fabrication capability since its introduction at the IBM Burlington fab in 1998.
- 1Q fidelities in lab 99.93% .
- 2Q fidelities in lab in non-iso-enriched Si of 92% ; anticipated that use of iso-enriched 28Si will allow 99%+.
- Strong coupling between SiGe quantum dot spins qubits and microwave photons demonstrated in lab, leading to potential of hybrid semiconductor-superconductor devices .
- Si/SiGe interface can be relatively well ordered, leading to fewer defects in comparison with SiMOS devices.
- Qubit spin information can ‘mix’ with other states known as ‘valleys’, due to small valley separation in SiGe, such leakage errors may be more of a concern here .
This technology is also expected to be able to operate at the relatively ‘mild’ temperature of 1K (though this has not yet been demonstrated). Multiple groups around the world (in particular academic efforts, for example at Princeton) are pursuing this technology, with commercial efforts led by Intel in partnership with QuTech.
Germanium has more recently emerged as an alternative platform for spin qubits. In this material ‘holes’ become practical to control. Such holes (the absence of a conduction electron) benefit from very small effective mass leading to highly isolated qubit states. 2Q gates have been demonstrated and 2D dot arrays realised . While germanium is the newest player in the field, its progress over the last two years is striking. We can expect devices with increased qubit counts from leading groups such as QuTech.
Silicon with Implanted Donor Spin in MOS devices
Individual atoms, commonly Phosphorus, are placed into the purified silicon substrate using ion implantation. The impurity appears positively charged when replacing a silicon atom in the lattice. This attraction confines the qubit electron. The donors are controlled by metallic gates on top of a Si/SiO2 dielectric, similarly to MOS quantum dots.
Notable Commercial Players: Photonic Inc.
- 1Q gates in lab 99.95% fidelity for electron spin; 99.99% fidelity for nuclear spin .
- Long qubit state lifetime (T1) 10s ; record values of coherence lifetimes (T2) 0.55s for electron spin, 35s for nuclear spin .
- Demonstrated entanglement of electron and nucleus with high fidelity .
- Different donor species can be implanted, including high-spin donors where the nucleus can be controlled by electric fields , or double donors with interesting optical properties (Morse et al. 2015).
- Fabrication is identical to silicon-MOS quantum dots, with the extra step of ion implantation, which is itself a standard CMOS process.
- Ion implantation imposes an intrinsic straggle ~10 nm in the final position of the donors. Requires 2Q gate designs that are insensitive to the precise inter-donor distance.
- Extreme cryogenic cooling to 20mK likely to be required for readout. Operation at higher temperature is known to be possible, from ensemble experiments.
Vancouver startup, Photonic Inc., is developing donor-based hardware with a focus on optical access.
Silicon with STM-fabricated donor-cluster Spin
Phosphorus donor atoms are placed with near-atomic precision on a silicon surface, using hydrogen depassivation lithography (HDL) and a scanning tunnelling microscope (STM). Both the control gates and the spin-carrying donors are fabricated with this method, and encapsulated within the crystalline silicon.
Notable Commercial Players: SQC
- Ultra-fast 800ps 2Q √SWAP gate demonstrated .
- Long spin state lifetime (T1) 30s in a 3P donor cluster .
- Current gate fidelity is thought to be limited by charge noise that scales with the ‘temperature’ of the electrons involved. Anticipated improvements via fabrication process and lower measurement/initialisation temperatures.
- Fabrication uses STM hydrogen depassivation lithography (HDL) to place individual donor atoms; this is well beyond established CMOS processes, but is in its own right an emerging future tech. Typically yields few-donor clusters, not single donors.
- No coherence control of a single qubit demonstrated so far.
- Extreme cryogenic cooling to 20mK likely to be required.
SQC is targeting building a 10Q prototype by 2023. The Australians are unlikely to have this space to themselves. Based in the US, Zyvex Labs pioneer atomically precise manufacturing technology and clearly see a much wider role for HDL across quantum technologies as a way to combat fabrication variability. Speaking at IQT New York, John Randal Zyvex CEO emphasised “traditional nanofabrication is an analogue process, quantum technology needs digital precision”.