Chapter 10. Major Qubit Platforms

In Chapter 9, we learned about five strict conditions (Devinzenzo’s criteria) required to build a useful quantum computer. The core challenge is resolving a fundamental dilemma: qubits must be sufficiently isolated from external noise to have long coherence times ($T_1, T_2$), while also being precisely controllable and measurable when needed.

Currently, research labs and companies worldwide are competing to create the “best qubits” based on different physical systems to resolve this dilemma. In particular, superconducting circuits, ion traps, neutral atoms, semiconductor spins, photons, and defect centers (such as NV centers) are the main candidates. This chapter compares the basic principles, recent performance trends, and pros and cons of the most promising quantum hardware platforms as of 2025.

1. Key Elements of Qubit Platforms

For any physical system to be used as a qubit, it must implement the following three core functions:

Definition (Encoding) of $ $ and $ $
Two quantum states (typically two energy levels) that are relatively stable against external noise but clearly distinguishable must be selected as $ $ (ground state) and $ $ (excited state). Depending on the physical implementation, various options exist, such as the direction of an electron’s spin, hyperfine levels, energy levels of a resonator, polarization of photons, or spin states of defect centers.
Control (Control – Gate)
To implement arbitrary rotations on the Bloch sphere, external electromagnetic pulses (microwaves, lasers, RF, etc.) with a frequency $ $ that exactly matches the energy difference $ E = $ between $ $ and $ $ must be applied. In practical devices, the amplitude, phase, and duration of these pulses are precisely controlled at the nanosecond scale (for superconducting/spin qubits) to the microsecond scale (for ion/neutral atom qubits), enabling the implementation of single- and two-qubit gates.
Interaction (Interaction – CNOT and other two-qubit gates)
Effective interactions between qubits must be designed so that the state of one qubit can influence the state of another. This is implemented through common electromagnetic resonators, collective vibrational modes (phonons), Rydberg interactions, exchange interactions, or nonlinear optical processes. According to Devinzenzo’s criteria, the key is whether these interactions are sufficiently fast (short gate time), accurate relative to noise, and allow for a flexible connection structure (e.g., 2D lattice, global coupling).

Another important point is the distinction between “physical qubits” and “logical qubits.” Since dozens to hundreds of physical qubits are required to construct a single logical qubit for error correction, the scalability of the platform has become as crucial as actual performance.

2. Comparison of Major Qubit Platforms (2025)

Today, quantum hardware considered as “serious candidates” in industry and research is generally categorized into five families: superconducting circuits, ion traps, neutral atoms, photons, and semiconductor spins. Defect centers such as diamond NV centers are increasingly specialized for sensor, communication, and node applications.

This section maintains the three platforms from the previous chapter (superconducting, ion traps, semiconductor spins) but briefly updates each platform’s representative performance metrics and research directions for 2023–2025.

Platform 1: Superconducting Qubits

System
Uses electrical circuits made from superconductors cooled to near absolute zero (about $10\,\mathrm{mK}$), typically consisting of LC resonators and Josephson junctions. Transmons and fluxonium are the most commonly used variants.
Definition of $ , $
A simple LC circuit behaves as a harmonic oscillator with energy $E_n \propto (n+1/2)$, resulting in equal spacing between $E_{0\to1}$ and $E_{1\to2}$, which causes easy leakage from $ $. Inserting a Josephson junction makes the potential nonlinear, altering the energy level spacing ($E_{0} E_{1}$), allowing selection of two levels to define $ $ and $ $.
Control (Gate)
A microwave pulse corresponding to $ E_{01} $ is applied to transmission lines on the chip to implement single-qubit rotation, with gate times of 10–50 ns being typical in current industrial standards.
Recent studies have reported energy lifetimes $T_1$ of transmons at the level of 0.3 ms and Ramsey coherence times of fluxonium reaching 1.5 ms through material (e.g., tantalum) and process improvements.
Interaction (CNOT)
Two qubits are weakly coupled via capacitors or resonators (e.g., coplanar waveguide resonators), and 2-qubit gates are implemented using effects such as exchange interaction or cross-resonance at specific frequencies.
Recent Performance Snapshot (Approximate Range)
- Coherence time: Based on transmons, $T_1, T_2$ are in the range of tens to hundreds of µs, with top devices reporting up to 0.1–0.3 ms.
- 1- and 2-qubit gate fidelity: Cases achieving 99.9% fidelity have been reported (particularly for fluxonium-type devices).

Pros	Cons
Fast gate speed $ _{} 1050, $, favorable for digital quantum algorithms.	$T_2$ is typically in the range of tens to hundreds of µs on most chips, requiring thousands or more physical qubits for large-scale error correction.
Prototypes with hundreds to thousands of qubits integrated on a single chip using photolithography-based semiconductor processes already exist.	Dilution refrigerators at ultra-low temperatures are essential, and complex wiring and electronic stacks make system-wide scalability a challenge.
Strong interaction design is easily achievable, enabling implementation of 2D lattice structures and various coupling topologies.[4][3]	Device-to-device variation and analysis of noise sources are difficult due to microscopic defects in Josephson junctions and material interfaces.

Platform 2: Trapped Ion Qubits

System
Ions (primarily Ca$^{+}$, Yb$^{+}$, etc.) are suspended in air within a vacuum chamber using electric fields or RF fields in a Paul trap or Penning trap, and are trapped in a line or 2D array.
Definition of $ , $
Qubits are defined by selecting two hyperfine or electronic levels of the ion, with frequency stability comparable to atomic clocks, making all ions effectively identical “replica qubits.”
Studies from the 2020s have reported exceptionally long coherence times for hyperfine qubits, ranging from 0.2–600 s.
Control (Gate)
Laser pulses focused on individual ions or ion pairs or microwave fields are used to rotate internal levels, with single-qubit gate times typically in the range of a few to several tens of µs.
Interaction (CNOT)
Ions repel each other via Coulomb forces and share collective vibrational modes (phonons) as a “bus,” transferring spin information from one ion to phonons and then recombining them with the spin of another ion using methods such as the Mølmer–Sørensen gate to implement two-qubit gates.
Recent Performance Snapshot
- Coherence time: Internal state $T_2$ reported up to seconds–hundreds of seconds.
- Gate fidelity: Results approaching 99.9% for both single- and two-qubit gates have been reported, and are also being utilized in error correction experiments.

Pros	Cons
Very long coherence time (seconds–minutes) and high measurement accuracy make it one of the cleanest physical qubits today.	Gate speed is relatively slow due to laser control and phonon mode adjustment, at the level of $ _{} 10100, $.
In small systems, it provides an interaction structure close to all-to-all connectivity, which is beneficial for compilation.	To scale to hundreds or thousands of ions, complex scaling strategies such as shuttling architectures and inter-module optical coupling are required.
Based on atomic physics, there is minimal device-to-device variation, and once calibrated, long-term reproducibility is good.	Experimental equipment is complex and bulky, requiring additional engineering for rack-mounted commercialization, including vacuum chambers, multiple lasers, and stabilization devices.

Platform 3: Semiconductor (Spin) Qubits

System
Gate electrodes are used to create a small cage called a “quantum dot” within semiconductor heterostructures such as silicon (Si/SiGe), and the spin of an electron (or nuclear spin) inside the quantum dot is used as a qubit.
$ , $ Definition
Under an external magnetic field $B$, the electron spin down $ $ and spin up $ $ states are defined as $ $ and $ $, respectively. Spin is less sensitive to noise than charge, allowing for fundamentally long coherence times.
Control (Gate)
In traditional ESR (Electron Spin Resonance), microwaves corresponding to the spin resonance frequency are applied to rotate the spin. For a more environmentally friendly approach, electrically induced spin resonance (E-DSR) is also being studied.
In the latest silicon quantum dot chips, single-qubit gate times are in the range of ns–tens of ns, and two-qubit exchange gates have demonstrated fast operation in the range of tens of ns–hundreds of ns.
Interaction (CNOT)
By placing two quantum dots tens of nm apart to allow their wavefunctions to overlap, exchange interaction occurs, and this is used to construct two-qubit gates via $ZZ$ or SWAP-like interactions.
Recent Performance Snapshot
- Coherence time: Spin $T_2$ reported up to ms–tens of ms in isotope-purified $^{28}$Si environments, which is among the best in solid-state platforms.
- Gate fidelity: Reports indicate that single-qubit fidelity above 99% and two-qubit CZ gate fidelity above 90% have been achieved in natural Si. | Pros | Cons | | :— | :— | | Can leverage existing CMOS semiconductor processes, theoretically enabling the highest level of scalability potential with millions of qubits integrated on a single chip. | Due to atomic-scale process variations, nuclear spin noise, and charge noise, actual device characteristics vary significantly, making calibration difficult. | | Can arrange qubits with very small pitch, naturally forming a 2D grid based on proximity interactions. | Device design, manufacturing, and control are still in the research phase, and there are many engineering challenges in large-scale simultaneous control and readout circuit integration. | | Spins themselves have long coherence times and high fidelity control, making them theoretically viable candidates that can meet error correction thresholds. | Being very close together leads to severe “crosstalk” and thermal management issues, requiring system-level design. |

3. Two Emerging Leading Platforms: Neutral Atoms and Photons

Since the mid-2020s, neutral atoms (Rydberg) and photon-based platforms have been evaluated as leading candidates on par with superconducting and ion trap platforms, rather than being classified as “others.” Neutral atoms, in particular, excel in analog/digital quantum simulation by arranging hundreds to thousands of atoms in 2D/3D configurations, while photon-based platforms offer unique advantages in room-temperature operation and networking, emerging as leading candidates for quantum internet and data centers.

Platform 4: Neutral Atom Qubits (Neutral Atom / Rydberg Qubits)

System
Cool neutral atoms (rubidium, cesium, etc.) with lasers and trap them in optical lattices or optical tweezers to create 1D/2D/3D arrays, using each atom as a qubit.
$ , $ Definition
Select the ground state and a long-lived excited state (or two hyperfine levels) to define the qubit, with the atoms themselves being “replicable” ideal systems similar to ions.
Control (Gates)
Use lasers to selectively excite and control specific atoms, with single-qubit rotations implemented via nanosecond–microsecond pulses.
Interaction (CNOT and Multi-Qubit Gates)
Exciting atoms to the high-energy Rydberg state causes their electron clouds to expand significantly, enabling strong interactions with other atoms at distances of hundreds of nm–µm. Due to Rydberg interactions and the “Rydberg blockade” effect, multi-qubit gates involving more than two qubits can be relatively naturally implemented.
Recent Performance and Scaling
- Number of Qubits: Quantum processors that can simultaneously control hundreds of atoms are already available in research and commercial cloud environments, with roadmaps targeting 1000 qubit scales.
- Coherence and Gates: Gate speed and fidelity are slightly lower than ion traps, but analog quantum simulators are already being applied to practical physical problems.

Pros	Cons
Can freely arrange hundreds to thousands of atoms in 2D/3D, making it one of the leading platforms in terms of “large-scale qubit count.”	Rydberg gate fidelity and stability are still lower than superconducting and ion platforms, and error accumulation management for long-running algorithms is difficult.
Excels in multi-qubit interactions and analog quantum simulation due to the Rydberg blockade.	Optical trapping and laser control are complex, and experimental and engineering challenges such as atomic loss and reloading issues exist.
Atoms themselves can be used in near-room-temperature environments, allowing operation without cryogenic refrigeration.	Single and two-qubit gate speed and fidelity are still in a transitional phase of rapid improvement, so long-term dominance is not yet certain.

Platform 5: Photonic Qubits

System
Uses photons (light) as qubits, encoding information via polarization $\lvert H\rangle, \lvert V\rangle$, time bins, paths, or continuous variables (phase, amplitude), etc.
Implements quantum operations using beam splitters, phase modulators, attenuators, etc., on integrated optical circuits (silica, silicon, SiN, etc.).
$ , $ Definition
For example, the presence of a photon in one of two paths can be defined as $\lvert 0\rangle$, and the absence in the other as $\lvert 1\rangle$, or various encoding schemes can be used, such as utilizing the two polarization states.
Control (Gates)
Combines linear optical elements (beam splitters, phase shifters, etc.) and measurements to implement measurement-based quantum computing (MBQC), or designs two-qubit gates using weak nonlinearity.
Interaction (CNOT)
Photons essentially interact very weakly, making it difficult to achieve strong two-qubit interactions in the classical sense.
Instead, research and development are ongoing to create large-scale cluster states and perform computations via measurements, or to construct “logical photons” using high-quality light sources, detectors, and switches.
Recent Trends
The photonic platform is evaluated as a highly attractive long-term candidate due to its capability for room-temperature operation, natural integration with communication networks, and modular scaling (“quantum data center”).
Indeed, several companies are aiming to realize logical qubits using large-scale photonic cluster states and error-correction-friendly GKP (bosonic) codes.

Pros	Cons
Can operate at room temperature and naturally integrate with optical fiber and communication infrastructure, making it optimal for quantum networks and cloud services.	Due to challenges such as photon loss, detection efficiency, and obtaining high-quality light sources, error correction for large-scale universal quantum computing is extremely difficult.
Inherently robust against decoherence and easy to maintain quantum states over long-distance transmission.	Strong two-qubit interactions are difficult to achieve, so most implementations rely on measurement-based or probabilistic gates.
A modular scaling roadmap is proposed, involving replicating and connecting chips and optical modules.	So far, there are no examples of stable large-scale logical qubit implementations, making the timing for practical algorithm execution uncertain.

4. Summary of Other/Complementary Platforms [4][3]

The following platforms are generally viewed as complementary candidates with strengths in specific applications (sensing, network nodes, hybrid architectures, etc.) rather than as “mainstream” platforms in 2025.

Platform	Qubit Encoding	Control Method	Pros	Cons
Diamond NV centers and defect centers (NV in diamond, NV in SiC, etc.)	Electronic and nuclear spin states in nitrogen-vacancy defects in diamond or SiC lattices.	Microwave, laser, magnetic field pulses.	Can maintain coherence times of tens of µs to milliseconds at room temperature, making it very promising for quantum sensors and quantum communication nodes.	It is difficult to strongly entangle distant defects or arrange them in large-scale lattice configurations.
Topological qubits (Majorana, etc.)	Information encoded in topologically protected quasi-particles (Majorana modes, etc.).	Theoretically proposed braiding operations, electric gates, etc.	Provides “self-error-correcting” qubits robust against defects in principle, allowing for high error tolerance.	Experimental verification is still lacking, and achieving reproducibility at the process and device level required for large-scale processors remains a challenge.

(Note: Neutral atom and photonic platforms have already been covered in a separate section in Chapter 3, so this table only summarizes defect-center and topological platforms.)

5. Practice Problems

Platform Selection (Update)
What are the platforms that are “the fastest but have relatively short coherence times” and “the slowest but have the longest coherence times”? Based on 2025 standards, describe where neutral atom platforms stand between these two extremes.
Josephson Junction and Nonlinearity
Explain why Josephson junctions are essential in superconducting qubits, from the perspective of “harmonic oscillator energy level structure” and the nonlinearity “$E_{0\to1} \neq E_{1\to2}$”, and describe how this contributes to improving gate fidelity.
Bus in Ion Trap CNOT
What is the physical degree of freedom that mediates information between qubits when performing a CNOT gate in ion trap qubits, and explain why this degree of freedom provides “global” coupling across the entire chain.
Semiconductor vs Superconducting – Scalability Discussion
Summarize the reasons why semiconductor companies like Intel are strategically showing significant interest in semiconductor (spin) qubits instead of superconducting qubits, from the perspectives of (a) utilization of existing CMOS processes, (b) chip integration density, and (c) long-term error correction.
Neutral Atom vs Ion Trap
Although both neutral atom qubits and ion trap qubits use atoms, they differ significantly in (a) trapping methods, (b) interaction mechanisms (phonon vs Rydberg), and (c) scalability strategies. Compare and contrast the two platforms for each of these three aspects.
Advantages and Disadvantages of Photonic Qubits
Although photonic qubits are highly attractive for operation at room temperature and communication, explain why they have not become the mainstream of universal quantum computers, from the perspectives of (a) two-qubit interactions, (b) loss and detection efficiency, and (c) error correction.

6. Explanation

Fastest/Slowest, and the Position of Neutral Atoms
The leading platforms that currently provide the fastest gates are superconducting qubits and semiconductor spin qubits, with gate times of approximately 10–50 ns (superconducting) or ns–several tens of ns (spin). Conversely, the slowest but most stable platform is ion traps, where gates take several–several tens of µs, but internal state coherence times can be on the order of seconds to minutes, which is highly favorable for reducing error accumulation. Neutral atom platforms lag slightly behind ion traps in terms of gate speed and fidelity, but they can arrange many more atoms in 2D/3D configurations, placing them in an intermediate position as “slower but providing a large number of qubits.”
Role of Josephson Junctions
A pure LC resonator is a harmonic oscillator with uniform energy level spacing, so microwave pulses targeting the $ $ transition simultaneously induce $ $ transitions, causing leakage to higher energy levels. Josephson junctions provide a nonlinear inductance, creating a non-harmonic potential, resulting in $E_{0\to1} \neq E_{1\to2}$, allowing selective resonance between only two levels. By separating energy spacing, control pulses rarely induce unwanted high-level transitions, reducing leakage errors and significantly improving the fidelity of one- and two-qubit gates.
Bus for Ion Trap CNOT: Collective Vibration (Phonon)
Ions are arranged in a line due to Coulomb repulsion, sharing collective vibrational modes like a single “elastic string.” By appropriately combining lasers, the internal spin state of a specific ion can be temporarily stored in this collective vibrational mode (phonon) and then recombined with the spin of another ion, transferring information via spin–phonon–spin. Since this collective vibrational mode spreads across the entire chain, ions can interact through the same phonon mode even at long distances, effectively functioning as a bus close to a fully connected structure.
Semiconductor vs Superconductor – Why Focus on Semiconductor Spin
Semiconductor spin qubits can leverage existing silicon-based CMOS processes, allowing the relatively direct use of current semiconductor industry infrastructure (300 mm wafers, process equipment).
Superconducting qubits are also fabricated using photolithography, but their materials, processes, and packaging differ from conventional logic CMOS, requiring the entire system, including cryogenic, high-frequency wiring, to be redesigned from scratch.
From the perspective of semiconductor companies, the vision of utilizing decades of accumulated expertise in fine processes and integrated circuit design to integrate hundreds of thousands of spin qubits “like transistors” provides significant long-term scalability advantages.
Neutral Atoms vs Ion Traps
In terms of trapping methods, ion traps use electric fields to capture charged ions, while neutral atoms use optical tweezers and lattices to hold charge-free atoms within the potential of light.
Regarding interaction mechanisms, ions use collective vibrations (phonons) caused by Coulomb interactions as a bus, whereas neutral atoms directly implement two-qubit or multi-qubit gates by utilizing strong long-range interactions generated when excited to Rydberg states.
In terms of scalability, ion chains face difficulties beyond tens to hundreds due to the complexity of phonon modes and the increasing number of control lasers, whereas neutral atoms can geometrically scale up to hundreds to thousands via 2D/3D arrays, although the quality and control precision of each qubit are still under development.
Why Photonic Qubits Are Challenging
Photonic qubits operate at room temperature and can transmit states over long distances with minimal loss through optical fibers, making them ideal for communication and networking.
However, photons naturally do not interact, requiring nonlinear media, measurement-based protocols, or large-scale auxiliary photons for two-qubit gates, leading to complex circuits and lower success probabilities.
Additionally, factors such as photon loss, incomplete detection, and non-ideal light sources can accumulate, exponentially increasing the resources (number of photons, modules) required for error correction, thereby delaying the practical realization of a universal photonic quantum computer.