Quantum Hardware Platforms
From superconducting circuits to topological qubits -- the physical implementations driving quantum computing
4.1 DiVincenzo Criteria
In 2000, David DiVincenzo identified five requirements that any physical system must satisfy to serve as a viable quantum computer:
- Scalable physical system with well-characterized qubits -- identifiable two-level quantum systems that can be scaled to large numbers
- Ability to initialize qubits -- reliably prepare a known initial state (e.g., $|0\rangle^{\otimes n}$)
- Long decoherence times -- much longer than gate operation times; the ratio $T_2/t_{\text{gate}}$ determines how many operations can be performed
- Universal set of quantum gates -- ability to implement arbitrary unitary operations through a discrete gate set
- Qubit-specific measurement -- ability to measure individual qubits in the computational basis with high fidelity
For quantum communication, two additional criteria are needed: the ability to interconvert between stationary and flying qubits, and faithful transmission of flying qubits.
4.2 Superconducting Qubits
Superconducting qubits are the leading platform, used by IBM, Google, and many others. They exploit the macroscopic quantum behavior of superconducting circuits cooled to ~15 mK.
The Transmon Qubit
The transmon (transmission-line shunted plasma oscillation qubit) is the most widely used superconducting qubit design. Its Hamiltonian is:
$$\hat{H} = 4E_C(\hat{n} - n_g)^2 - E_J \cos\hat{\varphi}$$
where $E_C$ is the charging energy, $E_J$ is the Josephson energy,$\hat{n}$ is the number of Cooper pairs, and $\hat{\varphi}$ is the superconducting phase. The transmon operates in the regime $E_J/E_C \gg 1$ (typically 50-100), which exponentially suppresses charge noise sensitivity while maintaining sufficient anharmonicity.
Transmon Parameters (State of the Art)
- Qubit frequency: $\omega_{01}/2\pi \approx 4\text{--}6$ GHz
- Anharmonicity: $\alpha/2\pi \approx -200\text{--}-300$ MHz
- Coherence times: $T_1 \approx 100\text{--}500\;\mu$s, $T_2 \approx 50\text{--}300\;\mu$s
- Single-qubit gate time: ~20-50 ns (fidelity $> 99.9\%$)
- Two-qubit gate time: ~30-100 ns (fidelity $> 99.5\%$)
- Readout fidelity: $> 99\%$
Two-Qubit Gates
Common two-qubit gate implementations in superconducting systems:
- Cross-resonance gate (IBM): Drive one qubit at the frequency of its neighbor. The ZX interaction creates a CNOT-equivalent gate.
- Tunable coupler (Google): A third transmon between two data qubits mediates an effective $iSWAP$ or $CZ$ interaction.
- Parametric gates: Modulate the flux through a tunable element to activate interactions at specific frequency differences.
Scalability and Challenges
- - Wiring bottleneck: each qubit requires multiple microwave control lines routed from room temperature to 15 mK
- - Frequency crowding: as qubit count grows, avoiding parasitic interactions becomes harder
- - Connectivity: typically limited to nearest-neighbor on a planar lattice
- - Current scale: IBM Eagle (127 qubits), Google Willow (105 qubits), IBM Condor (1,121 qubits)
4.3 Trapped-Ion Qubits
Trapped-ion quantum computers use individual atomic ions confined in electromagnetic traps. Qubits are encoded in long-lived electronic states of ions such as $\,^{171}\text{Yb}^+$, $\,^{40}\text{Ca}^+$, or $\,^{137}\text{Ba}^+$.
Qubit Encodings
- Hyperfine qubits ($\,^{171}\text{Yb}^+$):$|0\rangle = |F=0, m_F=0\rangle$, $|1\rangle = |F=1, m_F=0\rangle$. Splitting ~12.6 GHz. Extremely long coherence times ($T_2 > 10$ minutes).
- Optical qubits ($\,^{40}\text{Ca}^+$):$|0\rangle = 4S_{1/2}$, $|1\rangle = 3D_{5/2}$. Connected by a narrow optical transition at 729 nm. Coherence limited by the metastable lifetime (~1.2 s).
Gate Mechanisms
Single-qubit gates are implemented with focused laser beams or microwave pulses. Two-qubit gates use the shared motional modes of the ion chain:
$$\hat{H}_{\text{MS}} = \Omega \sum_{i} \hat{\sigma}_x^{(i)} (\hat{a} e^{-i\delta t} + \hat{a}^\dagger e^{i\delta t})$$
The Molmer-Sorensen gate drives a bichromatic laser field detuned by $\pm\delta$ from the motional sideband, creating an effective spin-spin interaction $\hat{\sigma}_x^{(i)}\hat{\sigma}_x^{(j)}$while disentangling the motion at the end of the gate.
Trapped-Ion Performance
- Single-qubit gate fidelity: $> 99.99\%$
- Two-qubit gate fidelity: $> 99.9\%$ (best demonstrated)
- Coherence times: seconds to minutes
- Connectivity: all-to-all within a chain (major advantage over superconducting)
- Gate speed: single-qubit ~1-10 $\mu$s, two-qubit ~10-200 $\mu$s (slower than superconducting)
- Current scale: Quantinuum H2 (56 qubits), IonQ Forte (36 qubits)
Scaling Approaches
- - QCCD architecture: Shuttle ions between zones in a segmented trap (Quantinuum approach)
- - Photonic interconnects: Entangle ions in separate traps via emitted photons
- - 2D trap arrays: Move beyond linear chains to 2D ion crystal configurations
4.4 Photonic Quantum Computing
Photonic quantum computers use photons as qubits, encoded in polarization, path, or time-bin degrees of freedom. Key advantages include room-temperature operation and natural compatibility with quantum networking.
Encoding Schemes
- Polarization:$|0\rangle = |H\rangle$ (horizontal), $|1\rangle = |V\rangle$ (vertical). Easy single-qubit gates via waveplates.
- Dual-rail (path):$|0\rangle = |10\rangle$ (photon in mode a), $|1\rangle = |01\rangle$ (photon in mode b). Beam splitters implement single-qubit gates.
- Squeezed states (continuous variable):GKP encoding or cat states in bosonic modes. Xanadu's approach.
The KLM Protocol
Knill, Laflamme, and Milburn (2001) showed that linear optics plus single-photon sources and detectors suffice for universal quantum computing. The key insight: photon-photon interactions (needed for entangling gates) can be simulated probabilistically using measurement and feedforward. A CZ gate succeeds with probability 1/16, but this can be boosted via teleportation gadgets to near-deterministic operation.
Measurement-Based Quantum Computing
An alternative paradigm: prepare a large entangled cluster state, then perform computation by single-qubit measurements alone. The measurement angles determine the gates. This is equivalent in power to the circuit model. PsiQuantum and Xanadu pursue this approach.
Photonic Platforms
- PsiQuantum: Silicon photonic chips with integrated sources, circuits, and detectors. Targeting 1M+ qubits.
- Xanadu: Borealis processor (216 squeezed-state modes). Demonstrated quantum advantage in Gaussian boson sampling.
- USTC (China): Jiuzhang processor achieved boson sampling advantage with 76-144 photons.
4.5 Topological Qubits
Topological quantum computing encodes information in non-local degrees of freedom that are inherently protected from local noise. This is the "holy grail" of quantum hardware -- achieving fault tolerance at the physical level.
Majorana Fermions
The leading approach uses Majorana zero modes -- exotic quasiparticles that are their own antiparticle ($\gamma = \gamma^\dagger$). A pair of Majorana modes encodes a fermionic mode, and the fermion parity of separated Majorana pairs stores quantum information:
$$\hat{c} = \frac{\gamma_1 + i\gamma_2}{2}, \quad |0\rangle_L: n = 0, \quad |1\rangle_L: n = 1$$
Since the information is encoded non-locally (in the joint parity of spatially separated Majorana modes), local perturbations cannot cause errors. Gates are performed by braiding -- physically exchanging the positions of Majorana modes:
$$\gamma_i \leftrightarrow \gamma_j: \quad \gamma_i \to \gamma_j, \quad \gamma_j \to -\gamma_i$$
Braiding Majorana modes produces gates in the Clifford group. For universality, additional non-topological operations (magic state distillation) are still needed.
Microsoft's Approach
- - Semiconductor-superconductor nanowires (InAs/Al, InSb/Al) hosting Majorana zero modes
- - In 2025, Microsoft announced the Majorana 1 processor -- the first topological qubit chip
- - Promise: inherently lower error rates, potentially reducing overhead for error correction by orders of magnitude
- - Challenge: creating and manipulating Majorana modes with sufficient quality remains at the frontier of condensed matter physics
4.6 Platform Comparison
4.7 Noise and Decoherence
All quantum hardware suffers from decoherence -- the loss of quantum coherence due to unwanted coupling with the environment. The key timescales are:
- $T_1$ (energy relaxation): Time for the qubit to decay from $|1\rangle$ to $|0\rangle$. Governed by $\rho_{11}(t) = \rho_{11}(0) e^{-t/T_1}$.
- $T_2$ (dephasing): Time for loss of phase coherence. $\rho_{01}(t) = \rho_{01}(0) e^{-t/T_2}$. Always $T_2 \leq 2T_1$.
- $T_2^*$ (inhomogeneous dephasing):Includes slow fluctuations. $T_2^* \leq T_2$. Can be partially refocused with echo sequences.
The quantum volume metric (IBM, 2019) provides a single-number benchmark accounting for qubit count, connectivity, and gate fidelity:
$$\text{Quantum Volume} = 2^n \quad \text{where } n = \max\{m : \text{can run } m \times m \text{ random circuits faithfully}\}$$