Silicon-based Quantum Computing

WOMANIUM Global Quantum Media Project Initiative — Winner of Global Quantum Media Project

Published in

Quantum Engineering

19 min readJul 20, 2023

01) Introduction
02) Classical Silicon Nanoelectronics
03) Moore’s Law: Shrinking Transistors
04) From Bits to Qubits
05) Quantum Silicon Nanoelectronics
06) Challanges
07) Types of Silicon Spin Qubits
08) Current Status of Silicon Spin Qubits
09) Fabricating Quantum Dots
10) Fabricating donors in Silicon
11) Fabricate the Qubit
12) Control the spins
13) Readout the spins
14) Storing quantum information
15) Donor spin qubit coupling mechanisms
16) Coupling with a shared electron: Wavefunction sharing
17) Exchange-coupled donor electrons: 10nm — 20nm regime
18) Electric dipole-coupled flip-flop qubits
19) Scale-up inwards: Higher spin nuclei
20) Timed Implementation: Not scalable
21) Scale up outwards: Deterministic Implantation
22) Demonstrated Deterministic Implantation
23) Deterministically implanted donor qubits
24) Future hardware developments
25) Future Applications
26) Conclusion

Introduction

Silicon’s role in classical nanoelectronics shaped the digital era with transistor-powered modern computers. Quantum effects hinder computation accuracy as transistors shrink. Quantum computing emerges as a cutting-edge solution to unleash unparalleled computational potential. Silicon’s abundance, scalability, and advanced fabrication techniques make silicon-based quantum computing intriguing. This article explores silicon-based donor qubits, investigating scalability through coupling mechanisms. It highlights silicon’s advantages while addressing material development and reproducibility challenges. The future promises billions of qubits on a chip, transforming industries and driving breakthroughs in diverse fields, igniting the quantum revolution in the industrial era.

Classical Silicon Nanoelectronics

Before delving into the construction of a silicon-based quantum computer, it is crucial to acknowledge that all current electronic devices are fabricated using silicon. Silicon is a semiconductor, making it an ideal material for constructing transistors — the fundamental building blocks of modern computers. Under normal conditions, silicon is electrically insulating and does not conduct electricity. However, through the process of doping — introducing impurities — it becomes possible to create metallic sources and drain electrodes. By applying a gate voltage to this device, the flow of electrons between silicon channels can be controlled. This manipulation switches the transistor from an off state to an on state, allowing current to flow when a voltage is applied. Thus, silicon transitions into an electrically conducting state. Additionally, silicon exhibits the ability to grow high-quality oxide when heated with oxygen, enabling the creation of electrically insulated gates for doped source and drain electrodes. Moreover, silicon is highly abundant, readily available, and cost-effective, making it an ideal material for computer manufacturing due to its abundance in the Earth’s crust — being the second most abundant element after oxygen.

Moore’s Law: Shrinking Transistors

Moore’s Law, predicting the exponential growth of computational power, is facing limitations as transistors shrink to nanometer scales. At such tiny dimensions, quantum effects, such as electron tunneling between closely spaced gates, become significant, rendering transistors unable to be in a completely off state. Consequently, this introduces errors into computations. To increase computational power beyond the limitations of classical transistors, exploration into different physics is necessary. Quantum computing, enabled by billions of transistors on a single chip, offers tremendous computational capabilities.

From Bits to Qubits

Classical computers utilize bits, representing 0 or 1 states based on transistor on and off states. In contrast, quantum computers utilize qubits, which are quantum systems with two energy levels — zero and one states. Qubits can exist in a superposition of both states simultaneously. While n classical bits contain only n bits of information, entangling n qubits allows for 2^n bits of information — an exponentially larger number. This exponential relationship empowers quantum computers with their exceptional computational prowess.

Quantum Silicon Nanoelectronics

To create a quantum system with two energy levels for qubits, one can leverage the spin of an electron in a magnetic field — an approach commonly referred to as a textbook qubit. By shrinking classical transistors down to a single electron in the transistor channel, we can harness the advantages of silicon to create these types of qubits.

Advantages of Silicon

1.Silicon offers several reasons for its selection as the material for building quantum computers. Its wide adoption in existing computer technologies has provided extensive knowledge and expertise in this domain. Utilizing scalable classical technology, such as that developed over the past 50 years for microelectronics, we can advance spin qubits. This enables the routine production of billions of identical transistors on a fingertip-sized area, facilitating the fabrication of qubits using similar technologies and materials, potentially accelerating quantum computing progress by decades.

2.Silicon qubits boast small footprints, measuring just 100nm by 100nm on a chip, making them particularly suitable for constructing two-dimensional qubit arrays. This opens the possibility of fitting millions of qubits on a single chip inside a conventional cryostat without requiring additional infrastructure, thus achieving a compact and fault-tolerant quantum computing processor.

3.Silicon’s possession of zero spin isotopes, such as silicon-28, creates a noise-free environment for qubits, providing long quantum states with low noise — a desirable trait not found in three-five semiconductors like gallium arsenide. Additionally, silicon’s weak spin-orbit coupling contributes to increased coherence time, making it one of the most coherent solid-state systems found in nature. Thanks to these factors and fast spin readout techniques, silicon qubits achieve single and two-qubit logic gates with fidelities above 99%, allowing for future incorporation of error-correcting algorithms for fault-tolerant computing. While silicon qubits typically require cooling to extremely low temperatures (tens of milli kelvin), promising research suggests they may be capable of operating at higher temperatures above 1K, which significantly simplifies cooling requirements when controlling millions of qubits.

The two-qubit device layout includes quantum dots D1 and D2 under gates G1 (blue) and G2 (red). Gates CB, G3, and G4 (purple and grey) form confinement barriers, and RG (yellow) supplies electrons. Gate electrodes ST, SLB, and SRB (green) create a single-electron transistor for charge sensing, while the AC current in the ESR line (light blue) manipulates electron spins. The control path involves initializing Q1 and Q2 by loading spin-down electrons into D2 (I1 and I2). Quantum operations are performed using ESR pulses in region. Q2 is read out via spin-dependent tunneling, and Q1 is transferred from D1 to D2 within 5 µs to avoid relaxation. Finally, Q1 is read out at the (0,0)-(0,1) transition, completing the operational sequence. Image Credits: [2]

Challanges

Despite the promise of silicon-based quantum computing, significant challenges remain before scaling up to larger numbers of qubits. Material development is a critical area of concern. To reliably produce qubits with low noise and low variability, isotopically enriched Si-28 transistor channels are essential to create the desired noise-free environment. High-quality interfaces devoid of charge caps or large fixed oxide charges are necessary for uniform qubits across the chip.

FIG. 1: Physical quantum processor. (a) SOI wafer with isotopically enriched silicon-28 forming the 2D qubit array and top silicon for qubit-operating transistors. Connections made through oxide regions using polysilicon vias. (b) Electrical circuit for controlling one Q-gate and one J-gate, enabling individual, row-by-row, or global operations. Physical architecture for one unit module with 480 qubits. Each J gate and qubit interconnected through the circuit in (b). Image Credits: [3]

Creating large-scale arrays of high-fidelity qubits for fault-tolerant architectures requires ingenious architectural designs, including dense 2D arrays of spin qubits in silicon with individually addressable qubits, interspersed control and readout electronics, and high connectivity to enable entanglement. The challenge lies in achieving reproducibility, as most progress in silicon qubits has been made in academic cleanrooms with limited process control and reproducibility. To produce industrial-scale arrays of qubits, a cleanroom with levels of reproducibility akin to the microelectronics industry is imperative.

Types of Silicon Spin Qubits

Silicon-based quantum computing encompasses three primary types of Quantum Dots, which are structures that confine electrons in silicon using surface nanoelectronics.

1.The first type is the Silicon MOS (Metal-Oxide Semiconductor) Quantum dots. Within these silicon MOS qubits, electrons reside in the silicon layer near the oxide interface, being electrically insulated from metallic gates by the oxide layer.

2.The second type of quantum dot architecture involves Silicon (Si)/ Silicon-Germanium (SiGe) heterostructures, where electrons inhabit the silicon well that is sandwiched between layers of silicon and germanium. These electrons are insulated from metallic gates positioned above them through an oxide layer.

3.The third type of silicon qubits comprises donors in Silicon. To produce these qubits, a single silicon atom is replaced with a donor qubit, which binds an additional electron through the coulomb potential to the donor nucleus. Notably, these qubit systems represent a “two-for-one” configuration, as they utilize both the spin of the electron and the spin of the donor nucleus as qubits.

Current Status of Silicon Spin Qubits

The present state-of-the-art for silicon spin qubits and the notable achievements thus far can be summarized as follows:

1. Number of Fully Operational Qubits: Currently, the number of qubits that have been successfully demonstrated in a Quantum Processor is 6, arranged in a linear array of Si/SiGe quantum dots. While this number may seem relatively small compared to platforms like superconducting systems, photons, and ion traps, where hundreds of qubits have been achieved, the future of silicon quantum computing appears promising due to upcoming collaborations with industry partners.

2. Coherence Time: The coherence time, a crucial metric for quantum information storage, has displayed impressive results for silicon spin qubits, particularly for Si-28 as a host. In a phosphorus donor spin qubit, record coherence times of approximately half a second have been achieved for the electron spin and over 30 seconds for the nucleus spin. These durations represent the record spin coherence times for single qubits in the solid-state domain.

3. Single Qubit Operation Time: An essential aspect is the speed at which qubits can be driven for applying gates. The current record for single qubit operation time is 42 nanoseconds, as measured for Si MOS nanowire quantum dots. This demonstrates the remarkable swiftness of gate operations in silicon-based quantum computing.

4. Single Qubit Gate Fidelity: Gate fidelity is a percentage indicating the accuracy of gate operations without errors. Si/SiGe quantum dots hold the record for single qubit gate fidelity, achieving an impressive 99.6%.

5. Two-Qubit Gate Performance: Efforts have pushed the two-qubit gate operation time down to 40 nanoseconds in Si/SiGe quantum dots, highlighting significant progress in the speed of two-qubit gate operations. Furthermore, two-qubit gate fidelities have been demonstrated to reach an exceptional 99.81%.

A false-colored scanning electron microscope image of a device similar to the one used in the experiments. Different colors represent various metallization layers. Quantum dots are defined by plunger (blue) and barrier (green) gates in the channel between screening gates (red). Additionally, two cobalt micromagnets (yellow) are positioned on top of the gate stack. Image Credits: [4]

Fabricating Quantum Dots

Fabricating silicon-based quantum dots involves depositing an isotopically enriched Si-28 layer or SiGe heterostructure on a handle wafer using chemical vapor deposition. An electrically insulating oxide layer is then formed through thermal oxidation or atomic layer deposition. A multilayer gate stack is created via electron beam lithography and metal deposition, enabling precise nanoscale patterning. Each layer is insulated by a native oxide formed when aluminum is exposed to air or by depositing a dielectric material using atomic layer deposition. The electron spin qubit is confined within the quantum dot devices by tuning the electrical potential of surface gates, attracting electrons beneath the gate structure and creating a controlled pool at the single electron layer level.

The images show scanning electron microscope views of devices with corresponding band bending diagrams and gate stacks. Dotted lines in (a-c) indicate cross-sections through the dot channel illustrated in (d-f), respectively. Crossed boxes indicate gates that overlap with implanted regions for ohmic contact. The plunger gates (yellow), barrier gates (blue), and screening gates (red) define the quantum dots. (a) SiMOS triple quantum dot linear array with two SETs for charge sensing. (b) Si/SiGe quintuple quantum dot linear array with two SETs for charge sensing. Ge/SiGe (2x2) quadruple quantum dot array, each tunnel-coupled to a metallic lead (green). Bandstructure of metal, dielectric (black), and semiconductor for (d) SiMOS, (e) Si/SiGe, and (f) Ge/SiGe. Image Credits: [5]

Fabricating donors in Silicon

In silicon-based quantum computing, donors are fabricated by substituting a single silicon atom in group four of the periodic table with a group five donor atom. The donor nucleus, having an extra positive charge compared to the surrounding silicon, binds an additional electron through the coulomb potential. This eliminates the need for surface nanoelectronics used to confine electron spins in quantum dots. The system provides a two-for-one qubit configuration, as it encompasses both the spin of the nucleus and the electron. For instance, using phosphorus as a donor results in a spin one-half system, offering spin up and spin down states. Fabrication methods for these donor spins involve two primary approaches.

1.The first method employs a scanning tunneling microscope (STM) tip to pattern the silicon surface using a high hydrogen resist. Specific hydrogen atoms are locally removed, facilitating the incorporation of phosphorus donor atoms through phosphine gas at desired sites defined by the STM. The phosphorus donors are then encapsulated by growing additional silicon, allowing for subsequent control and readout for the qubit system.

2.The second method utilizes ion implantation to introduce donor atoms into silicon. This approach enables mass selection from various group five donors, including arsenic and bismuth, thereby exploring more exotic physics. Moreover, ion implantation is compatible with standard semiconductor industry fabrication methods, making it a scalable pathway. While the following paragraphs will focus on a specific worked example, the fundamental principles are applicable to most spin qubits in silicon.

**STM technique (left) and ion implantation(right)** Image Credits: [1]

Fabricate the Qubit

The fabrication of a silicon-based quantum computing qubit begins with a pristine silicon wafer. The initial steps involve growing an oxide layer on the surface and implanting a single donor atom into the silicon. This implantation process employs a method using an iron-based firing system to introduce the single atom. Subsequently, nanoscale electronics are fabricated onto the surface by patterning aluminum gates using electron beam lithography. These gates serve the crucial role of controlling and reading out the spin qubits in future experiments.

Once the qubit device is fabricated, it undergoes packaging and connection to the outside world, allowing electrical signals to be applied to the gates for qubit control and readout. The device is wire bonded to a printed circuit board, enabling the application of electrical signals. The complete setup is mounted onto a dilution refrigerator capable of cooling the device down to ultra-low temperatures around 10mK, significantly colder than outer space. At these temperatures, the donor electron becomes bound to the donor nucleus. External magnetic fields generated by a large solenoid magnet surrounding the refrigerator are applied to Zeeman split the electron and nucleus spins, leading to a spin Hamiltonian that describes their interactions.

Electron Zeeman Interaction:

Nuclear Zeeman Interaction:

Hyperfine Interaction:

Control the spins

Controlling the spins in the system involves using a microwave antenna, indicated in green. The microwave antenna applies an oscillating magnetic field capable of spinning the spins around. The first step is to ensure that the donor electron remains trapped onto the donor nucleus by raising the Fermi level of the single electron transistor (SET) island above the energy levels of both the spin-up and spin-down states of the donor qubit. This prevents the electron from tunneling onto the single electron transistor during the experiment. Then, the microwave antenna is used to apply an oscillating signal, resulting in two electron spin resonance peaks corresponding to the nuclear spin of the donor being either spin-down or spin-up. This flips the electron spin in both cases because the microwave frequency matches the energy splitting between the electron’s spin-down and spin-up states. The same on-chip microwave antenna can also be used to control the nuclear spins by applying different frequencies.

Qubit device and pulsing scheme. (a) Scanning electron micrograph of the qubit device with the SET and donor. (b-c) Pulse sequence for qubit initialization, control, and readout. (d) Energy level diagram of the 31P electron-nuclear system. (e-f) Microwave pulse sequence and PL gate voltage waveform for spin manipulations. (g) Example of ISET response to consecutive read/control events at B0 = 1.07 T, indicating spin tunnel-in and tunnel-out times. Image Credits: [6]

Readout the spins

To read out the state of the qubit, determining whether it’s a zero or a one, the electron spin is coupled to a SET defined with surface gates on the quantum device. The SET acts as a highly sensitive charge detector. To operate the readout system, the Fermi level of the SET island is tuned to lie between the spin-up and spin-down states of the donor electron. When the electron is spin-up, it has enough energy to tunnel off the donor and onto the SET island, causing the donor to become ionized and carry a relative positive charge due to the loss of the electron. This results in a current flowing through the SET, leading to a “current blip.” However, when the electron is spin-down, it lacks the required energy to tunnel onto the SET island, keeping the current in a low state. By observing the presence or absence of current in the SET, the electron spin state (zero or one) can be effectively determined.

Spin readout device configuration and charge transitions. (a) Schematic of spin-dependent tunneling configuration, allowing only spin-up electrons to tunnel onto the SET island. (b) Pulsing sequence for single-shot spin readout and corresponding SET response, ISET. Scanning electron micrograph of a similar device with marked P donors implant area. (d) SET current ISET as a function of voltages on the top and plunger gates. Charge transfer events disrupt the SET Coulomb peaks. (e) Line traces of ISET showing shifts in SET current peaks caused by ionizing the donor. The charging energy of the SET is EC ~ 1.5 meV. Image Credits: [7]

Storing quantum information

In silicon-based quantum computing, ensuring the long-term storage of quantum information is essential for meaningful control experiments without decoherence. To achieve this, isotopically purified silicon, specifically silicon-28, is utilized to avoid magnetic noise introduced by silicon-29’s nuclear spin. Silicon-28 provides an ideal low noise environment for qubits, often referred to as a “semiconductor vacuum.” This is achieved by depositing an epitaxial layer of silicon-28 onto a natural silicon handle wafer, where the donor qubits reside, resulting in significantly increased coherence times compared to natural silicon. Coherence times of up to 30 seconds have been achieved for nuclear spins in Si-28, ensuring high-fidelity gate operations and minimizing decoherence errors before performing gate operations.

a) Scanning electron micrograph image of a device similar to Device A, showing the P donor, microwave (MW) antenna, and the SET for spin readout. b) Schematic of the Si substrate, comprising an isotopically purified 28Si epilayer (with 800 ppm residual 29Si) on a natural Si wafer.c) Energy level diagram of the coupled e−-31P0 system (left) and the ionized 31P+ nucleus (right). Quantum states are encoded using oscillating magnetic fields B1 at frequencies νe1,2 ≈ γeB0 ± A/2 for electron spin resonance (ESR) and νn1,2 ≈ A/2 ± γnB0 for nuclear magnetic resonance (NMR). The ionized 31P qubit is operated at the frequency νn0 = γnB0.Image Credits: [8]

Donor spin qubit coupling mechanisms

While single donor qubits have been thoroughly explored and optimized, enabling quantum algorithms requires implementing two-qubit gate operations, necessitating the coupling of two donor qubits. Depending on the length scale, several coupling mechanisms are employed. At scales below 10nm, donors can be coupled through single electron sharing between two donor nuclei. At slightly larger distances, around 10–20nm, each donor possesses its own electron, allowing coupling through electron exchange interactions. Longer length scales, around 100nm or more, involve the creation of flip-flop qubits, combining nuclear and electron spins. Electric dipole interactions between these flip-flop qubits enable larger donor spacings. Future possibilities include using intermediate quantum dots to shuttle electrons between donors, facilitating micron-scale distances and offering more control and readout options on a chip. The focus will be on exploring the first three coupling mechanisms for donor spin qubits.

Coupling with a shared electron: Wavefunction sharing

One method of coupling donor qubits involves two nuclear spins sharing a common electron. By implanting two phosphorus nuclei closely spaced (around 6 nanometers apart), the donors share the same electron, resulting in an electron wavefunction that overlaps both nuclei. As a consequence, the single electron resonance spectrum exhibits four different resonance frequencies, corresponding to the four spin combinations of the two nuclei (down-down, down-up, up-down, up-up). Moreover, the nuclear magnetic resonance spectrum displays two peaks, allowing the extraction of individual hyperfine couplings between each nucleus and the shared electron. This shared electron enables the implementation of a two-qubit control Z gate between the two nuclei. By applying a two-pi pulse on the electron, conditional rotations of the nuclear spins are achieved, enabling quantum operations on this coupled system.

Operating a one-electron — two-nuclei quantum processor. (a) Illustration of a pair of 31P nuclei (red), asymmetrically coupled to the same electron (blue). On-chip oscillating magnetic fields (yellow) control the spins. (b) Effective-mass calculation of the electron wavefunction ψ(y, z) on the 2P cluster, reproducing observed hyperfine couplings. Experimental NMR and ESR spectra, yielding hyperfine couplings A1 ≈ 95 MHz and A2 ≈ 9 MHz between the electron and nuclear qubits Q1, Q2. (d) Implementation of a geometric two-qubit CZ gate, where a conditional π phase shift is acquired on the electron spin based on nuclear spin states. This operation corresponds to the CZ gate on the nuclei within the electron |↓⟩ subspace. Image Credits: [9]

Exchange-coupled donor electrons: 10nm — 20nm regime

In this configuration, two phosphorus donors are implanted approximately 20 nanometers apart. At this distance, each donor nucleus binds its own donor electron, and these two electrons are exchange coupled. By magnetically detuning the electrons using hyperfine interaction, two distinct electron spin resonance transitions are observed. In the weak exchange regime, a native CNOT gate is implemented using a single electron spin resonance pulse, which performs conditional rotations on the target electron depending on the state of the control electron. Two electron spin resonance transitions implement the CNOT gate, while two other transitions serve as zero-CROT gates, resulting in various two-qubit gate possibilities.

The device consists of a two-qubit metal-oxide-semiconductor system. (a) Scanning electron micrograph of a similar device used in the experiment, with labeled aluminum gates. (b) Schematic cross-section showing a pair of donors approximately 10 nm below a thin SiO2 dielectric within an isotopically-enriched 28Si epilayer. The gate setup enables control of the electron spins through energy detuning provided by opposite nuclear spin states. Experimental observations have been in the regime J ≥ 100 MHz, preventing the use of a specific two-qubit gate. A SWAP operation was recently demonstrated between strongly exchange-coupled electron spins but without coherent quantum control. Image Credits: [10]

Electric dipole-coupled flip-flop qubits

Examining larger length scales on the order of 100 nanometers or more, flip-flop qubits can be generated using anti-parallel spin states of both the nucleus and the electron. These flip-flop qubits experience electric dipole resonance transitions by applying an alternating bias on a gate above the donor, modulating the electron’s position relative to the donor nucleus. The hyperfine coupling between the donor electron and nucleus is affected by this modulation. In the weak exchange regime, a native CNOT gate can be realized through electron dipole spin resonance on resonance with the energy splitting between flip-flop states. The electric dipole interaction allows long-range coupling of flip-flop qubits, rendering them insensitive to precise donor qubit placement, which is particularly advantageous for ion-implanted donor qubits.

The flip-flop qubit utilizes the electron and nuclear spin states of a 31P donor. (A) Energy level diagram showing electron (↑, ↓) and nuclear (⇑, ⇓) spin states with oscillating magnetic fields inducing transitions. (B) The flip-flop qubit represented on a Bloch sphere. Device layout with a single-electron transistor (SET) for spin readout, local gate electrodes for control, and MW antennas for electric and magnetic control of donor spins. Orange for nuclear spin, blue for electron spin, and green for the flip-flop qubit. Image Credits: [11]

Scale-up inwards: Higher spin nuclei

For scalability, a potential approach involves moving to higher spin nuclei beyond phosphorus. Antimony-123, for instance, has a nuclear spin of seven halves, resulting in eight different nuclear energy levels instead of the two states (spin up and spin down) for phosphorus. These higher spin nuclei can be explored to operate as qudits, higher-dimensional systems, offering potential benefits for error correction or utilizing the system as multiple qubits. Additionally, higher spin donor nuclei possess a quadrupole moment, enabling the exciting possibility of electrically controlling the nucleus spin. An applied electric field can distort the nucleus’s charge distribution and drive nucleus electric resonance, potentially leading to the use of smaller electrical antennas for driving the spins instead of the current large on-chip microwave antennas.

The experiment uses a silicon metal-oxide-semiconductor device with 3Sb nuclear spin. The device has short-circuited antenna terminations with gaps. Shear strain in the silicon substrate is shown in a cross-section. The energy level diagram of the spin-7/2 nucleus of an ionized 123Sb donor displays d 7 individually addressable nuclear resonances, with the mI = −1/2 ↔ +1/2 transition forbidden in NER. Image Credits: [12]

Timed Implementation: Not scalable

While the timed implementation method has been valuable for exploring small numbers of donor qubits, it is not scalable for producing the large arrays needed for a functional quantum computer with error correction capabilities. The current approach relies on sprinkling a random distribution of a small number of donor ions into a silicon quantum device and then tuning the device to locate nearby qubits to a single electron transistor. However, to achieve large arrays of qubits, an alternative approach is required.

Scale up outwards: Deterministic Implantation

A promising strategy for scaling up involves deterministic implantation, which requires the realization of single ion detectors in the quantum devices. To achieve this, a pin diode is fabricated on the silicon chip with a reverse bias applied to the electrodes. When a donor ion is implanted into the detector’s construction site, the electric field gradient from the electrodes generates electron-hole pairs in the silicon, sweeping them to opposite electrodes in the detector. This induces a detectable ion beam induced charge signal for single ion implantation events, allowing the detection of each donor ion with high fidelity.

Demonstrated Deterministic Implantation

In an experimental demonstration, phosphorus ions were implanted into a single ion detector using a nano stencil atomic force microscope cantilever. By measuring the ion beam induced charge signal for each of the 10,000 detection events, a distribution of energy of the created electron-hole pairs was obtained due to the random nature of the donor ion stopping processes in the oxide and silicon. Through optimization of the noise threshold of the detector, a single ion detection fidelity of up to 99.85% for single phosphorus ions was achieved, providing high confidence in the deterministic implantation of donors and paving the way for scalable arrays of donor qubits.

The single-ion implanter system incorporates an AFM with a Peltier-cooled sample stage housing charge-sensitive preamplifier electronics. The AFM cantilever is configured as a stencil with a 16 μm diameter microaperture to localize the ion beam within a single construction site. In situ optical AFM camera view shows wire bonding connecting the single-ion detector top electrode to the preamplifier circuit board. The AFM image provides nanometer precision alignment between ion implant sites and device processing steps. Image Credits: [13]

Deterministically implanted donor qubits

With the capability of deterministically implanting single donor ions into silicon, the challenge now lies in integrating these ions into qubit devices with precise alignment. Developing a fabrication process flow to achieve this integration is underway, with the goal of creating deterministically implanted donor qubits within qubit devices. Once this milestone is achieved, the focus will shift towards scaling up to larger arrays of qubits to implement quantum algorithms effectively.

Future hardware developments

The future of hardware development in the field of silicon spin qubits will transition from the current academic era to the industrial era. The academic era, involving research conducted in university clean rooms, allows for the exploration of fundamental physics in spin qubits and the production of around ten qubits in silicon. However, this falls behind other quantum computing platforms in terms of qubit numbers. In the industrial era, large-scale companies with semiconductor foundry capabilities will leverage their expertise in fabricating reliable and consistent transistors to produce vast numbers of small qubits in silicon. This transition may pave the way for billions of spin qubits integrated onto a single silicon chip in the coming decades.

Future Applications

As silicon quantum computers scale up, they will have a transformative impact on various fields. The ability to simulate molecules and their interactions using quantum computers will revolutionize drug discovery and materials design, potentially leading to significant advancements in medicine and technology. Quantum computers will also play a crucial role in cryptography, offering enhanced security for online communications and data protection. Furthermore, the efficiency of quantum computers in solving linear equations will greatly benefit machine learning and artificial intelligence applications, leading to improved forecasting capabilities, such as weather prediction. Additionally, quantum computers’ capability to simulate interactions at the atomic level could unlock the secrets of nuclear fusion, providing access to clean and unlimited energy sources. By simulating chemical reactions, quantum computers may contribute to carbon capture and storage techniques, bolstering efforts to combat climate change.

Conclusion

Silicon has proven to be an excellent platform for building quantum computers due to its scalability and the wealth of knowledge from classical computers that can be leveraged. The advantages and challenges of silicon-based spin qubits have been explored, covering various coupling mechanisms, deterministic ion implantation, and potential hardware developments. While the current academic era focuses on exploring fundamental physics and small-scale qubit production, the transition to the industrial era holds promise for scaling up quantum computers with billions of qubits integrated on a single chip. The future applications of quantum computing in simulating molecules, advancing cryptography, improving machine learning, and addressing climate challenges indicate the tremendous potential of silicon-based quantum computing for transforming various industries and scientific fields.

References

[1] D. Holmes, “Hardware Lecture: Si-based Quantum Computing,” presented at WOMANIUM QUANTUM, Jul. 20, 2023. [Online]. Available: https://www.youtube.com/watch?v=pFP7RvD9HvQ.

[2] W. Huang et al., “Fidelity benchmarks for two-qubit gates in silicon,” Nature, vol. 569, no. 7757, pp. 532–536, May 2019. doi: 10.1038/s41586–019–1197–0.

[3] M. Veldhorst et al., “Silicon CMOS architecture for a spin-based quantum computer,” Nature Communications, vol. 8, no. 1, Dec. 2017. doi: 10.1038/s41467–017–01905–6.

[4] S. G. J. Philips et al., “Universal control of a six-qubit quantum processor in silicon,” Nature, vol. 609, no. 7929, pp. 919–924, Sep. 2022. doi: 10.1038/s41586–022–05117-x.

[5] W. I. L. Lawrie et al., “Quantum dot arrays in silicon and germanium,” Applied Physics Letters, vol. 116, no. 8, Feb. 2020. doi: 10.1063/5.0002013.

[6] J. J. Pla et al., “A single-atom electron spin qubit in silicon,” Nature, vol. 489, no. 7417, pp. 541–545, Sep. 2012. doi: 10.1038/nature11449.

[7] A. Morello et al., “Single-shot readout of an electron spin in silicon,” Nature, vol. 467, no. 7316, pp. 687–691, Sep. 2010. doi: 10.1038/nature09392.

[8] J. T. Muhonen et al., “Storing quantum information for 30 seconds in a nanoelectronic device,” Nature Nanotechnology, vol. 9, no. 12, pp. 986–991, Oct. 2014. doi: 10.1038/nnano.2014.211.

[9] M. T. Mądźik et al., “Precision tomography of a three-qubit donor quantum processor in silicon,” Nature, vol. 601, no. 7893, pp. 348–353, Jan. 2022. doi: 10.1038/s41586–021–04292–7.

[10] M. T. Ma{‘{c}}dzik et al., “Conditional quantum operation of two exchange-coupled single-donor spin qubits in a MOS-compatible silicon device,” Nature Communications, vol. 12, no. 1, Jan. 2021. doi: 10.1038/s41467–020–20424–5.

[11] R. Savytskyy et al., “An electrically-driven single-atom ‘flip-flop’ qubit,” arXiv:2202.04438 [quant-ph], 2023.

[12] S. Asaad et al., “Coherent electrical control of a single high-spin nucleus in silicon,” Nature, vol. 579, no. 7798, pp. 205–209, Mar. 2020. doi: 10.1038/s41586–020–2057–7.

[13] A. M. Jakob et al., “Deterministic Shallow Dopant Implantation in Silicon with Detection Confidence Upper-Bound to 99.85% by Ion-Solid Interactions,” Adv. Mater., vol. 34, p. 2103235, 2022. DOI: 10.1002/adma.202103235.

[14] “Roadmap,” Diraq, [Online]. Available: https://diraq.com/roadmap.