⚛️ Quantum Computing · Hardware
📅 Березень 2026⏱ 13 min🔴 Advanced

Qubit Technologies: Superconducting, Trapped Ion, and Photonic Qubits

Building a quantum computer requires maintaining quantum coherence — superposition and entanglement — against constant decoherence from thermal noise and the environment. Different physical implementations of qubits make radically different trade-offs between coherence time, gate speed, gate fidelity, and scalability.

1. DiVincenzo Criteria for Qubits

David DiVincenzo (2000) formulated 5 requirements any physical qubit system must satisfy to perform quantum computation:

  1. Scalable: A system of well-characterised qubits that can be scaled to large numbers.
  2. Initialisable: Ability to reset qubits to a known starting state |0⟩ with high fidelity.
  3. Long coherence: Decoherence times (T₁, T₂) much longer than gate operation time.
  4. Universal gate set: Ability to implement a universal set of quantum gates.
  5. Measurable: Ability to read out each qubit state with high fidelity.
Key metrics: T₁ = energy relaxation time (|1⟩ → |0⟩ probability decay, amplitude damping) T₂ = dephasing time (loss of phase coherence, |+⟩ → mixed state) T₂ ≤ 2T₁ always. Often T₂ << 2T₁ due to dephasing noise. Gate fidelity F = Tr(U†_ideal · ρ_actual) ∈ [0,1] F > 0.999 typically required for fault-tolerant error correction (surface code threshold: gate error < ~1%) Clock rate = 1/t_gate Effective circuit depth ≈ T₂ / t_gate

2. Superconducting Qubits (Transmon)

The dominant commercial platform (IBM, Google, Rigetti). A Josephson junction — two superconductors separated by a thin insulating barrier — creates a non-linear quantum inductor. Combined with a capacitor, this forms an anharmonic quantum oscillator:

Transmon Hamiltonian (charge basis): H = 4E_C (n̂ - n_g)² - E_J cos(φ̂) E_C = charging energy = e²/2C (Coulomb energy per electron pair) E_J = Josephson energy ∝ I_c (critical current) n_g = gate charge (offset from integer) φ = superconducting phase difference Transmon regime: E_J >> E_C (ratio ~50-100) → qubit frequency ω ≈ (8E_J E_C)^{1/2} / ℏ ≈ 4-8 GHz → anharmonicity α = ω₁₂ - ω₀₁ ≈ -E_C/ℏ ≈ -200 to -350 MHz (negative: |2⟩ is closer to |1⟩ than |1⟩ to |0⟩) → charge noise insensitive (unlike earlier CPB qubit) Operating conditions: Temperature: 10-20 mK (dilution refrigerator) Qubit frequency spacing: several MHz to avoid crosstalk Control: microwave pulses at ω through coplanar waveguide Readout: dispersive coupling to microwave cavity (150-250 MHz) Current state (2024): T₁: 100-500 μs (IBM Heron r2) T₂: 50-300 μs Single-qubit fidelity: 99.9-99.99% Two-qubit (CZ) gate fidelity: 99.0-99.9% Gate time: 20-100 ns (fast!) Qubit count: 100-1000 qubits on single chip

3. Trapped Ion Qubits

Individual ions (typically ⁴⁰Ca⁺, ⁸⁸Sr⁺, ¹⁷¹Yb⁺, or ¹³³Ba⁺) are levitated in Paul traps — oscillating electric fields — and cooled to their motional ground state with laser cooling. The qubit is encoded in two long-lived electronic levels:

Qubit encoding: ¹⁷¹Yb⁺: hyperfine ground states |0⟩=|F=0,m_F=0⟩, |1⟩=|F=1,m_F=0⟩ Transition frequency: ~12.6 GHz (microwave) Natural coherence time (no decoherence): years to decades(!) Actual T₂ in trap: seconds to minutes → limited by magnetic field noise, trap frequency instabilities, and laser frequency noise Gate operations: Single-qubit: laser/microwave pulse (Rabi oscillations) Two-qubit: Mølmer-Sørensen gate via shared motional bus - Apply bichromatic laser field to both ions - Excite/deexcite phonon modes → mediates ion-ion interaction - Gate time: ~10-500 μs (slow vs superconducting) Current state (2024): T₁: > hours (optical qubits) or minutes (hyperfine) T₂: 1-10 s (with dynamic decoupling) Single-qubit fidelity: 99.99% Two-qubit fidelity: 99.6-99.9% Gate time: ~100-500 μs (100× slower than SC) Qubit count: 30-50 ion qubits in linear trap (IonQ, Quantinuum) Hundreds: multi-zone or 2D trap arrays

4. Photonic Qubits

Photons are excellent qubits — they travel at the speed of light and almost don't interact with the environment (extremely long coherence). The challenge is making them interact with each other for two-qubit gates:

Photonic encoding options: Polarisation qubit: |H⟩ = |0⟩, |V⟩ = |1⟩ (λ/2 plate = single-qubit gate) Path/dual-rail qubit: photon in one of two paths Time-bin qubit: photon arrival in early/late time slot Two-qubit gates in linear optics: KLM scheme (Knill, Laflamme, Milburn 2001): Probabilistic gates using ancilla photons + photon detection + feedforward. Success probability per gate: O(1/n²) without tricks. With fast feedforward + percolation: scalable but resource-intensive. Continuous Variable (CV) photonics: Encode qumodes in Gaussian states (squeezed states of EM field). Deterministic gates via squeezing + beamsplitters. measurement-based quantum computing on cluster states. Current platforms: PsiQuantum: fault-tolerant photonic QC using silicon photonics foundry Xanadu (Borealis): 216-mode Gaussian Boson Sampling (2022, quantum advantage claim) Advantages: Room temperature (no dilution fridge), natural for networking (qubits as photons) Challenges: Low gate efficiency, deterministic photon sources still imperfect

5. Spin Qubits

Electron or nuclear spins in semiconductors (silicon, germanium, III-V compounds) or in defect centres in wide-bandgap materials (NV centres in diamond, silicon-vacancy SiV in diamond) serve as qubits:

Silicon spin qubits (Intel, QuTech): Qubit: electron spin in silicon double quantum dot |↑⟩ = |0⟩, |↓⟩ = |1⟩ in magnetic field B₀ Larmor frequency: f = g·μ_B·B₀/h ≈ 10-40 GHz at B₀=0.5-2T Exchange interaction two-qubit gate: J·(σ₁·σ₂) Heisenberg coupling → controlled by gate voltage: fast (~10-100 ns) when tuned Advantage: compatible with CMOS manufacturing Challenge: single-atom precision placement, hyperfine noise from ²⁹Si Solution: isotopically purified ²⁸Si (nuclear spin-0 removes dominant noise) 2024 state: T₂ up to seconds with ²⁸Si, fidelities >99%, ~6 qubit arrays NV centres in diamond: Electron spin of nitrogen-vacancy defect T₁ ~ ms at RT, T₂ ~ μs at RT, ~ ms at low T Single-photon emission → spin-photon interface → quantum networking Used in quantum repeaters, quantum sensing (magnetometry)

6. Topological Qubits

Topological qubits aim to store quantum information non-locally in a topologically protected state, making decoherence exponentially suppressed by the physical gap:

Majorana zero modes (Microsoft Station Q approach): Predicted to appear at ends of semiconductor nanowires (InAs, InSb) coupled to superconductors in a magnetic field. A qubit encoded in two Majorana modes separated over macroscopic distance: Local perturbations cannot distinguish |0⟩ from |1⟩ → inherently protected. Topological protection: Decoherence rate ∝ exp(-L/ξ) where L = separation, ξ = coherence length If L >> ξ: effectively zero decoherence from local noise. Gates: "Braiding" Majoranas (non-Abelian anyons) implements unitary gates. A braid group element = topologically protected gate operation. Current status (2025): Microsoft's topological qubit chip "Majorana 1" (2025 announcement) claims demonstration of "topological qubits" — still under independent verification. True fault-tolerant topological quantum computing remains a research goal.
The threshold theorem: Quantum error correction can suppress errors to negligible levels if physical error rates are below a threshold (~1% for surface codes). Superconducting and trapped ion platforms both approach this threshold. However, a fault-tolerant logical qubit requires ~1000 physical qubits per logical qubit depending on error rate. A useful fault-tolerant computer (e.g., Shor's algorithm on RSA-2048) needs ~4000 logical qubits = ~4 million physical qubits. No current system is close.

7. Technology Comparison

Qubit technology comparison (approximate, 2024-2025): ┌──────────────┬──────────┬──────────┬──────────┬──────────┬──────────┐ │ Platform │ T₂ │ 2Q fidel │Gate time │ Scale │ Temp │ ├──────────────┼──────────┼──────────┼──────────┼──────────┼──────────┤ │ Supercond. │ ~100 μs │ 99.5% │ ~50 ns │ 100-1k │ 10-20 mK │ │ Trapped Ion │ ~1-10 s │ 99.8% │ ~100 μs │ 30-100 │ RT │ │ Photonic │ ~μs-ms │ ~99% │ fast │ 200+mode │ RT │ │ Si Spin │~10 ms-1s │ >99% │ ~10 ns │ 6-16 │ 10-100mK │ │ NV diamond │ ~ms │ ~99% │ ~μs │ 1-10 │ RT │ └──────────────┴──────────┴──────────┴──────────┴──────────┴──────────┘ No single platform wins on all criteria. Superconducting: current leader in scale + speed Trapped ion: current leader in fidelity + coherence Photonic: natural quantum networking, room temperature Spin qubits: semiconductor manufacturing compatibility — future scalability Likely future: hybrid systems with specialised co-processors per task.