Defect and Solid-State Spin Qubits

An optically addressable spin defect is an atom-scale quantum system embedded in a macroscopic crystal. Its localized electronic levels can supply a controllable spin, its optical transition can connect that spin to a photon, and nearby nuclear spins can supply longer-lived storage. The same ingredients support three rather different technologies: a local sensor, a small hybrid processor, or a node in a photonic quantum network.

That compactness is the platform’s central opportunity and its central difficulty. The host is not an inert package. Isotopes, strain, surfaces, implantation damage, charge traps, phonons, and nanophotonic fabrication all enter the operational quantum model. A statement such as “the defect has a one-second coherence time” says little about a network unless it also states which spin, pulse sequence, optical activity, temperature, and device geometry produced that number.

This page connects the spin, nuclear register, and optical interface of a defect to a working memory or network node. It covers vacancy complexes, optically active silicon defects, and selected rare-earth ions. Silicon Spin Qubits discusses gate-defined dots and donor processors; NV-Center Sensing develops signal estimation with defects. Current experimental milestones are discussed in Quantum Hardware Scaling.

Required background. Hardware Overview introduces preparation, control, and measurement. Spin in Magnetic Fields and Hyperfine Structure explain Zeeman levels and electron–nuclear coupling.

From a Crystal Defect to a Qubit

Localized levels in a band gap

A point defect changes the crystal potential and can create localized electronic states inside the host band gap. Occupying those states with a particular number of electrons defines a charge state. Crystal-field and spin–orbit interactions then split orbital and spin multiplets. An optical transition between localized configurations makes the defect a color center when it absorbs or emits within the relevant spectral range.

The useful level structure is therefore conditional on more than chemical identity. It depends on charge state, local strain, electric field, isotope, crystallographic orientation, and sometimes magnetic field. Optical illumination can change the charge state, so ionization and charge recovery must be included when modeling preparation, measurement, and stored-state survival.

For a spin-$S$ ground manifold, a common effective Hamiltonian is

$$\frac{H}{\hbar} =\;D\!\left[S_z^2-\frac{S(S+1)}{3}\right] +E(S_x^2-S_y^2) +\gamma_e\mathbf B\cdot\mathbf S +\sum_k \mathbf S\cdot\mathbf A_k\cdot\mathbf I_k +\frac{H_{\rm strain}+H_{\rm electric}+H_{\rm nuc}}{\hbar}.$$

The spin matrices $\mathbf S,\mathbf I_k$ are dimensionless. $D,E,A_{k,ij}$ have angular-frequency units; the remaining $H$ terms have energy units and exclude effects already included in $D$ and $E$. Here $\gamma_e$ denotes the coefficient of the positive electron Zeeman term ($g_e\mu_B/\hbar$ in an isotropic model), not the negative signed gyromagnetic ratio defined through $\boldsymbol\mu=\gamma\mathbf S_{\rm physical}$. An anisotropic $g$ tensor requires the corresponding tensor Zeeman term. $D$ and $E$ describe axial and transverse zero-field splittings, while $\mathbf A_k$ is the hyperfine tensor for nuclear spin $k$. This is an effective Hamiltonian within a chosen manifold. Optical excitation may change all of these parameters, and group-IV vacancy centers require explicit orbital and spin–orbit degrees of freedom when their orbital branches are not safely eliminated.

Zero-phonon light and phonon sidebands

An optical excitation can return without changing the vibrational state or can create or absorb phonons. The former contributes to the zero-phonon line (ZPL); the latter produces a phonon sideband. Define the Debye–Waller fraction by

$$f_{\rm DW} = \frac{\Gamma_{\rm ZPL}} {\Gamma_{\rm ZPL}+\Gamma_{\rm PSB}}.$$

These are radiative rates for the specified environment; this ratio excludes nonradiative decay and is not the photon quantum yield. The standard defect properties are reviewed by Doherty and colleagues and Wolfowicz and colleagues. Coherent remote interference generally uses ZPL photons, so a small $f_{\rm DW}$ is a direct efficiency cost unless a cavity selectively enhances the useful channel. NV centers offer excellent spin coherence and mature control, but only a few percent of their free-space emission is in the ZPL. Several group-IV centers direct a substantially larger fraction into their ZPL and have inversion-symmetric structures that reduce first-order sensitivity to some electric-field fluctuations.

The trade is not free. Orbital branches can couple strongly to acoustic phonons, shortening spin coherence unless temperature, strain, or dressed-state control suppresses the process. “Better optical emitter” and “better memory” are separate claims.

Interface spins and memory spins

The localized electron spin is usually the fastest and most optically visible degree of freedom. A nearby host or impurity nucleus is slower but can remain coherent while the electron is manipulated or repeatedly reset. A useful node therefore assigns roles:

Physical degree of freedom	Typical role	Strength	Recurring cost
defect electron spin	communication qubit, ancilla, local sensor	fast microwave control and optical access	optical reset, charge noise, shorter memory under activity
intrinsic defect nucleus	local memory or flag	deterministic coupling and known identity	isotope availability and fixed hyperfine structure
nearby host nucleus	data or memory qubit	potentially long storage and several qubits per defect	device-specific coupling graph and slow gates
emitted photon	flying qubit or herald	long-distance transport	loss, indistinguishability, collection, detection
orbital or charge state	optical interface, auxiliary, or readout handle	strong coupling to light or electronics	leakage, ionization, phonon sensitivity

Calling every row a qubit without stating its role obscures the architecture. The photon is not a resident register qubit, and a nuclear memory may be usable only through the electron ancilla.

Representative Platform Families

Nitrogen-vacancy centers in diamond

The negatively charged NV center has an electronic spin-triplet ground state. At low magnetic field its axial zero-field splitting is approximately $D/(2\pi)=2.87\ \mathrm{GHz}$. A qubit commonly uses $m_s=0$ and one of the $m_s=\pm1$ levels. Microwaves drive the electron spin, optical pumping initializes it, and spin-dependent fluorescence or resonant optical cycling provides readout.

The electron couples to the native nitrogen nucleus and to nearby $^{13}\mathrm C$ nuclei. This creates a hub-and-spoke register: the electron is the optically accessible ancilla or communication qubit, while selected nuclei are data, memory, or flag qubits. Dynamical-decoupling sequences on the electron can both protect coherence and synthesize conditional nuclear rotations. Direct radio-frequency driving expands the accessible coupling range but adds timing and calibration demands.

NV centers support room-temperature coherent control and sensing. The high-fidelity resonant optical operations used for single-shot readout and remote entanglement normally require cryogenic operation. Room-temperature spin control therefore does not imply a room-temperature network node.

Group-IV vacancy centers in diamond

Silicon-, germanium-, and tin-vacancy centers place a group-IV impurity between two vacant lattice sites. Their approximate inversion symmetry suppresses a leading electric-dipole response to uniform electric fields and can yield stable, narrow optical transitions. The commonly studied negative charge states have spin-$1/2$ electronic ground manifolds and large ZPL fractions, making them useful optical interfaces. Charge state is essential: neutral SiV has a different spin structure. A large ZPL fraction alone does not establish transform-limited emission or mutual photon indistinguishability.

For SiV centers, phonon-driven transitions between orbital branches can limit spin coherence at ordinary cryogenic temperatures. Millikelvin operation, large strain, or continuous dressing can reduce that channel. Heavier group-IV centers have larger orbital splittings and may relax the temperature demand, but fabrication yield, spectral uniformity, charge stability, and complete node demonstrations remain species- and device-dependent.

An isotopically selected $^{29}\mathrm{SiV}$ center contains a spinful silicon nucleus as a built-in memory. A $^{117}\mathrm{SnV}$ center similarly supplies an electronuclear register. These deterministic nuclei avoid searching for a particular nearby host isotope, although additional host nuclei can still extend the register.

Defects in silicon carbide

Silicon carbide combines a wide-band-gap host with mature wafer processing, multiple polytypes, and several optically addressable defect families. Neutral divacancies often have spin-$1$ ground states, while negatively charged silicon vacancies can provide spin-$3/2$ manifolds. The latter can be used as a qudit or restricted to a two-level subspace, in which case spectator transitions and leakage must be measured explicitly.

Falk and colleagues show how polytype and defect configuration change the available spins. Wafer processing is useful for integration, but optical linewidth, charge stability, and spin coherence must survive the processing used for the actual device.

T centers and rare-earth ions

The silicon T center combines an optically active electron spin with a telecom O-band transition and coupled nuclei. Its photon interface, rather than electrostatically defined confinement, distinguishes it from a gate-defined silicon dot or phosphorus donor processor.

Rare-earth ions such as $^{171}\mathrm{Yb}^{3+}$ are substitutional impurities with useful optical and hyperfine levels. Their inhomogeneous spectrum makes matching emitters difficult but can permit frequency multiplexing when individual transitions remain resolvable. Shielded electronic orbitals help some transitions remain narrow; host, site, isotope, temperature, and magnetic field still determine the actual coherence and selection rules.

Other emitters, including defects in hexagonal boron nitride, are important research directions. Their inclusion in a platform comparison should follow demonstrated charge-state control, reproducible spin assignment, coherent control, and a quantified optical interface, not brightness alone.

Sensing, local registers, and network nodes

A sensor

A sensor uses the electron spin to accumulate a signal-dependent phase or relaxation probability. Spatial resolution, contrast, collection rate, surface distance, and calibration matter alongside coherence. A nearby spin bath may be the signal rather than an error. These measurements are developed in NV Centers and Solid-State Defects and Magnetometry.

A local hybrid register

A register uses the electron as a controllable bus between nuclear memories. The control system must identify the coupling tensor of each chosen nucleus, compile conditional gates, protect spectators, and map nuclear states back to the electron for readout. Its capacity is therefore set by addressability and gate quality, not by the number of spins physically present.

A quantum-network node

A network node adds an optical interface, low-loss collection, photon filtering, frequency or phase stabilization, detectors, synchronization, and real-time logic. The communication spin is repeatedly optically reset while a nuclear memory stores earlier entanglement. Memory coherence must therefore be measured during network activity, not only while the optical interface is idle.

These three uses can share a physical defect yet need different figures of merit. Optimizing shallow implantation for sensing can worsen spectral diffusion. Etching a nanocavity can improve collection while degrading spin or optical coherence. An ensemble optimized for magnetic-field sensitivity does not automatically provide individually addressable network nodes.

Initialization and Readout

Optical pumping and resonant cycling

Spin-selective nonradiative decay can polarize a defect into a preferred spin state. Under resonant cryogenic excitation, resolved spin-conserving transitions can instead support repeated cycling and high-fidelity single-shot readout. Both mechanisms are dissipative. The laser changes the spin and can also change the charge state.

The full readout record may include phonon-sideband fluorescence, resonant reflection, arrival times, a charge-state signal, or a repeated mapping from a nucleus to the electron. A quoted assignment fidelity should state whether it is raw, corrected for preparation, conditioned on a charge check, or postselected on consistent repetitions.

For a binary result, report the response matrix

$$M= \begin{pmatrix} P(0|0) & P(0|1)\\ P(1|0) & P(1|1) \end{pmatrix},$$

not only its average diagonal. Readout asymmetry matters when a measurement outcome controls feed-forward.

Spin-to-charge and nuclear-assisted readout

Spin-to-charge conversion maps spin information onto a longer-lived charge configuration and then reads that charge optically or electrically. Repetitive nuclear-assisted readout maps the same memory state to the electron several times. Both can improve signal-to-noise ratio, but each added cycle creates opportunities for memory flips, ionization, and false heralds.

Initialization has the same accounting. A nuclear state may be prepared by measurement and feedback, by polarization transfer from the electron, or by selective optical pumping. A heralded preparation probability must not be reported as deterministic initialization fidelity.

Coherent Control and Local Gates

Electron-spin rotations

Microwave magnetic fields drive electron-spin resonance. In a rotating frame, a selected two-level transition has the familiar control Hamiltonian

$$\frac{H_{\rm rot}}{\hbar} = \frac{\Delta}{2}\sigma_z +\frac{\Omega(t)}{2}\cos\phi(t)\,\sigma_x +\frac{\Omega(t)}{2}\sin\phi(t)\,\sigma_y.$$

Here $\sigma_z$ assigns $+1$ to the chosen upper level, $\Delta=\omega_q-\omega_d$, and the transverse laboratory drive is $\hbar\Omega\cos(\omega_dt+\phi)\sigma_x$. The rotating frame is $\exp(-i\omega_dt\sigma_z/2)$, as derived in Silicon Spin Qubits. The two-level approximation requires resolved spectator transitions and weak drive relative to the carrier; leakage into other spin or orbital levels must be checked in an enlarged model. A fast Rabi frequency is valuable only with a measured error model.

Hyperfine-mediated nuclear control

In a secular approximation, one electron coupled to several nuclei can be written

$$\frac{H}{\hbar} =\;\omega_e S_z +\sum_k \omega_k I_{kz} +\sum_k A_{k,\parallel}S_zI_{kz} +\sum_k A_{k,\perp}S_zI_{kx} +H_{nn}/\hbar.$$

The $S,I$ matrices are dimensionless, all $\omega,A$ coefficients are angular frequencies, and $H_{nn}$ is an energy-valued nuclear interaction. The electron Zeeman/zero-field gap must justify neglecting terms that flip the electron. Each nucleus has its own transverse $x$ axis chosen along the transverse hyperfine vector; transverse nuclear precession is retained. The nuclear precession axis depends on the electron state. Alternating free evolution with electron $\pi$ pulses can therefore create an electron-controlled nuclear rotation. Direct radio-frequency pulses can act on nuclei while electron decoupling sequences preserve the interface spin. These gates are powerful but device specific because the $\mathbf A_k$ tensors depend on atomic position.

Two nuclear qubits can be entangled through a sequence of electron-mediated conditional gates, through their direct dipolar coupling, or by a geometric phase accumulated by an ancilla cycle. The electron should return disentangled from the data at the end. Residual ancilla entanglement is a coherent error, and reduces the purity or correlations of the retained nuclear state.

Local electron–electron coupling

Nearby defect electrons may couple through magnetic dipole interaction, exchange, a shared optical mode, or a mechanical mode. Dipolar coupling falls as $r^{-3}$ and competes with implantation uncertainty and spectral crowding. Cavity- or phonon-mediated proposals relax geometric locality, but a measured single-spin coupling is not yet a scalable two-qubit gate. Local-array claims should report the number of simultaneously controlled defects, the coupling graph, idle errors, and calibration overhead.

The Spin–Photon Interface

Photon probability is a product

The probability that one excitation produces a useful detector event can be factored into sequential conditional probabilities as

$$\eta_{\rm node} = p_{\rm prep}\, p_{\rm exc}\, \beta_{\rm mode}\, \eta_{\rm out} \times \eta_{\rm filt}\, \eta_{\rm conv}\, \eta_{\rm det}.$$

$\beta_{\rm mode}$ is the fraction emitted into the collected optical mode; $\eta_{\rm conv}$ is included when quantum frequency conversion is used. Each factor must refer to the same experiment and be conditioned on the preceding successful steps; independence is not needed for that chain rule. Extra photons or false clicks require separate branches. Different papers quote efficiencies at different physical locations. Multiplying the best component values from different devices does not produce a measured node efficiency.

Cavity rates and branching probability

Use $H_{\rm int}/\hbar=g(a^\dagger\sigma_-+a\sigma_+)$. Let $\kappa$ be the cavity photon-number decay rate, so an empty-cavity field amplitude decays at $\kappa/2$. Let $\gamma$ be the emitter excited-state population decay rate outside that cavity mode. With negligible pure dephasing, its optical coherence decays at $\gamma/2$. In this convention,

$$C=\frac{4g^2}{\kappa\gamma}.$$

On resonance in the fast-cavity limit $\kappa\gg g,\gamma$, eliminating the cavity amplitude gives an additional emission rate $\Gamma_c=4g^2/\kappa$. Competing decay channels then give

$$\beta_c=\frac{\Gamma_c}{\Gamma_c+\gamma} =\frac{C}{1+C}.$$

Here $\gamma$ includes unwanted ZPL emission, phonon-sideband emission, and nonradiative population loss. Since $g$ is the coupling of the particular ZPL transition, its oscillator strength already enters $C$; multiplying this $\beta_c$ by $f_{\rm DW}$ again would double-count that penalty. If the useful port contributes $\kappa_{\rm ex}$ to $\kappa$, the collected branch is $\beta_c\kappa_{\rm ex}/\kappa$ before later transmission loss. Strong coupling, detuning, or appreciable pure dephasing requires solving the actual cavity-emitter dynamics rather than applying this branching formula. See Cavity QED for the open-system model.

Indistinguishability and spectral diffusion

Remote interference requires photons to overlap in frequency, linewidth, polarization, arrival time, and temporal envelope. In a simple Markovian model with total excited-state population decay rate $\Gamma$ and optical pure dephasing rate $\gamma^*$, two independently emitted, otherwise identical single photons have integrated Hong–Ou–Mandel visibility

$$V_{\rm HOM} = \frac{\Gamma}{\Gamma+2\gamma^*}.$$

The assumptions are impulsive excitation, an exponential emission envelope, Markovian dephasing, and collection without temporal or spectral postselection. The normalized temporal density matrix is

$$\rho(t,t')=\Gamma e^{-\Gamma(t+t')/2} e^{-\gamma^*|t-t'|},$$

for $t,t'\ge0$, after removing the common carrier. Its purity $\operatorname{Tr}\rho^2$ gives the formula above. Slow diffusion, timing jitter, extra photons, and unequal wave packets change the measured visibility. Filtering or accepting only selected detection times can increase conditional visibility at a rate cost; the displayed value is not a universal upper bound under postselection. Kiraz, Atatüre, and İmamoğlu analyze the relation between source dynamics and indistinguishability. Stark tuning can align two emitters, but large tuning ranges may expose charge instability. Inversion symmetry helps only with the noise channels to which the transition is symmetry-protected.

Telecom conversion

Many defect transitions are outside the low-loss telecom windows. Quantum frequency conversion can translate their photons while preserving the encoded qubit. It adds conversion loss, pump-induced noise, filtering, phase control, and another calibration loop. For fibre attenuation $\alpha$ in $\mathrm{dB/km}$ over length $L$,

$$\eta_{\rm ch}=10^{-\alpha L/10}.$$

A wavelength described as “near telecom” still needs its actual attenuation and conversion decision stated. Native wavelength, converted wavelength, and deployed fibre length must be specified when calculating transmission.

Heralded Remote Entanglement

Single-click and two-click scaling

In a midpoint protocol, each communication qubit is entangled with an optical mode. Photons interfere, and a detector pattern heralds a remote spin state. Schematically,

$$|\psi\rangle_j = \sqrt{q}\,|0\rangle_j|1\rangle_j^{\gamma} +\sqrt{1-q}\,|1\rangle_j|0\rangle_j^{\gamma}.$$

Let $\eta$ be the probability that a photon emitted at either node reaches and is detected at the midpoint. In ideal symmetric protocols, one detected photon gives success scaling $\propto\eta$, while two required detections give $\propto\eta^2$. If nodes are separated by length $L$, each arm has fibre transmission $10^{-\alpha L/20}$; its square is the full-length transmission $10^{-\alpha L/10}$. “Linear versus quadratic” is ambiguous unless the transmission being varied is specified.

For the state above, suppose both nodes emit independently, photons are indistinguishable, phase is stable, both output ports are accepted with the appropriate local correction, detectors resolve photon number, and dark counts are absent. Accept exactly one detected photon. There are two branches:

$$\begin{aligned} p_{\rm desired}&=2q(1-q)\eta,\\ p_{\rm false}&=2q^2\eta(1-\eta). \end{aligned}$$

The first branch prepares a Bell state; the second comes from two emissions with one photon lost and leaves both spins in $|00\rangle$. Consequently

$$\begin{aligned} p_{\rm succ}&=2q\eta(1-q\eta),\\ F_{\rm Bell}&=\frac{1-q}{1-q\eta}. \end{aligned}$$

Reducing $q$ suppresses false heralds but slows successful preparation. Threshold detectors can also accept two photons in one output port, so they require a different expression. Two-round schemes such as Barrett–Kok use a second herald to remove unwanted spin branches; their loss scaling and phase-stability requirements follow the complete sequence.

For repeated independent attempts of duration $T_{\rm att}$ and success probability $p_{\rm succ}$,

$$\langle T_{\rm link}\rangle \simeq \frac{T_{\rm att}}{p_{\rm succ}}.$$

The distribution is geometric, so the mean is not a guarantee. Timeouts can protect memory fidelity but reduce yield.

Active-link efficiency

The relevant memory time is the coherence time while the communication qubit is being reset, excited, and measured. Define a dimensionless active-link efficiency

$$\eta_{\rm link}^{\rm active} = \frac{T_{\rm mem}^{\rm active}} {\langle T_{\rm link}\rangle} \simeq \frac{T_{\rm mem}^{\rm active}p_{\rm succ}} {T_{\rm att}}.$$

This ratio compares timescales for one sequential link attempt stream; it is neither a survival probability nor a universal repeater threshold. Multiplexing and error correction change the relevant waiting-time model. For independent geometric waiting $N\ge1$ and exponential coherence $e^{-NT_{\rm att}/T_{\rm mem}^{\rm active}}$, the average retained coherence is

$$\overline C= \frac{p_{\rm succ}e^{-x}} {1-(1-p_{\rm succ})e^{-x}},\qquad x=\frac{T_{\rm att}}{T_{\rm mem}^{\rm active}}.$$

This is not the coherence evaluated at the mean waiting time. A timeout or correlation between success and memory damage changes the conditional average. Entanglement fidelity, swapping, multiplexing, local gates, classical latency, and comparison with direct transmission are still required.

Store first, link second

A multi-link node must store one successful link while attempting another. Repeated optical reset of the electron changes the hyperfine field seen by a nuclear memory and can produce stochastic phase kicks. Charge-state failures may also alter the memory Hamiltonian. Dynamical decoupling, decoherence-protected subspaces, real-time phase tracking, and error-detecting measurements all address this problem. An idle $T_2$ measurement misses it.

Noise during storage and optical operation

Spin baths

Host nuclei and paramagnetic impurities create quasi-static and dynamical magnetic noise. Isotopic purification reduces the bath but can also remove candidate memory nuclei. A deliberately retained nucleus may be a data qubit; an uncontrolled nucleus at similar coupling may be crosstalk. The same atom can therefore move between resource and environment as control improves.

Phonons

Phonons drive spin-lattice relaxation, orbital transitions, and optical dephasing. Their effect depends strongly on level splitting and temperature. Cooling, strain engineering, phononic band structures, and dressed states can change the rate. Sukachev and colleagues demonstrate the importance of suppressing phonon-mediated orbital transitions for SiV spin storage.

Charge and electric-field noise

Nearby traps shift optical frequencies and can switch the defect charge state. Surface proximity and nanofabrication often amplify these effects. Charge repumping restores operation but adds dead time and can disturb stored spins. Spectral stability should therefore be reported over the full experimental duty cycle, not only during selected resonant scans.

Strain and orientation

Strain can split unwanted orbital degeneracies, tune transitions, and suppress phonon processes. It can also vary across a wafer and complicate spectral matching. Magnetic-field alignment affects selection rules and spin mixing. A result obtained at one carefully selected orientation is not automatically a fabrication-tolerant architecture.

Optical leakage and backaction

Off-resonant excitation, imperfect selection rules, intersystem crossing, ionization, and detector dark counts create unheralded errors or false heralds. Optical operations on the communication spin can dephase nearby memories. Error budgets should distinguish:

• erasure or heralded failure, where the run is discarded knowingly;
• assignment error, where the classical outcome is wrong;
• state error, where an accepted quantum state is wrong;
• leakage, where population leaves the encoded manifold;
• memory damage, where network activity corrupts stored information.

Heralding converts some loss into waiting time, but it does not convert every physical error into erasure.

Fabrication and Systems Engineering

Create the right defect in the right place

Defects can be native, grown in situ, implanted, irradiated and annealed, or created by focused laser processing. A scalable route must jointly control position, orientation, isotope, charge state, optical linewidth, and spin coherence. High placement precision with severe lattice damage is not a useful win. Post-implantation annealing and overgrowth can repair some damage but add process variation.

Put the emitter at the optical field maximum

Nanocavities and waveguides require spatial and dipole alignment between a single emitter and a small optical mode. Fabricate-then-find and find-then-fabricate workflows trade alignment accuracy against throughput. Pick-and-place integration can select good emitters but introduces assembly and packaging overhead. Yield must be quoted for the complete usable node, not for bright defects before cavity coupling.

Tune without destabilizing

Local electrodes, strain actuators, magnetic fields, and optical dressing can match transition frequencies. Each tuning channel consumes wiring, bandwidth, and calibration effort. A large static tuning range is not enough if the transition diffuses faster than the feedback loop can track it.

Package the whole node

A network experiment may need a cryostat, confocal or fibre coupling, microwave and radio-frequency delivery, narrow-linewidth lasers, frequency conversion, single-photon detectors, phase stabilization, time tagging, and FPGA feedback. Integrated photonics can shrink parts of this stack, but the external lasers, detectors, and refrigeration still contribute to power, latency, physical size, and operating cost.

Metrics That Survive Comparison

Report metrics at an explicit reference plane and operating mode:

Layer	Useful metrics	Essential qualifiers
spin	$T_1$, Ramsey $T_2^*$, echo and decoupled $T_2$	temperature, field, sequence, optical state, depth
local gates	Clifford or process fidelity, leakage, duration	simultaneous or isolated, register size, spectators
readout	full response matrix, duration, erasure fraction	charge conditioning, repetition, correction method
optical transition	linewidth, diffusion, lifetime, $f_{\rm DW}$	timescale, resonant power, device geometry
cavity interface	$g$, $\kappa$, $\gamma$, $C$, mode efficiency	convention, input/output port, spectral filtering
network link	success probability, Bell fidelity, attempt rate	fibre length, loss, conversion, phase control
memory	$T_{\rm mem}^{\rm active}$ or attempts to $1/e$	exact network activity and reset sequence
system	accepted-state rate, availability, recalibration time	postselection, timeout, duty cycle, number of nodes

Independent classical “no-fault” probabilities can be multiplied, but component quantum fidelities generally cannot. Coherent errors can cancel or reinforce, and a detected failure can select a different state ensemble. The node’s output channel or task success must therefore be measured or calculated under the actual preparation, conditioning, and timing.

Local correction and distributed gates

Electron–nuclear registers can use repeated parity checks, flagged ancillas, and small quantum codes. Abobeih and colleagues provide an example of fault-tolerant circuit constructions in a nuclear-spin register. A circuit construction restricts fault propagation under stated assumptions; demonstrating lower logical error additionally depends on physical noise, decoder performance, and repeated operation.

Remote entanglement, local gates, and classical feed-forward can teleport a gate between modules. Accepting every teleportation measurement outcome after a usable entangled pair has been supplied makes the conditional gate deterministic. It does not make the preceding optical pair generation deterministic. The complete elapsed time includes failed attempts and memory decay while waiting. Quantum Teleportation develops the protocol, while Modular Architectures connects it to scheduling.

Exercises

1. Spin-one transition frequencies

For

$$\frac{H}{\hbar}=DS_z^2+\gamma_e B_zS_z,$$

find the $m_s=0\rightarrow m_s=\pm1$ angular frequencies. State the condition under which this simplified expression is trustworthy.

Solution

The level energies in angular-frequency units are

$$\frac{E_m}{\hbar}=Dm^2+\gamma_eB_zm.$$

Therefore

$$\begin{aligned} \omega_{0\to+1}&=D+\gamma_eB_z,\\ \omega_{0\to-1}&=D-\gamma_eB_z. \end{aligned}$$

These are positive excitation frequencies only while $|\gamma_eB_z|<D$. Outside that regime the level differences change sign and the corresponding spectroscopic frequencies are their absolute values. The formula assumes a field aligned with the defect axis and neglects transverse strain, electric fields, hyperfine structure, and level mixing. It is not a general vector-field calibration formula.

2. Conditional hyperfine phase

Take an electron–nuclear interaction

$$\frac{H_{\rm hf}}{\hbar}=A S_zI_z$$

with electron eigenvalues $s_0$ and $s_1$. How long must the system evolve to produce a relative nuclear phase $\pi$ between the two electron branches for a nuclear eigenvalue difference $\Delta m_I=1$?

Solution

The branch-dependent angular frequency difference is

$$\Delta\omega =A(s_1-s_0)\Delta m_I.$$

The accumulated relative phase is $\Delta\phi=\Delta\omega t$. Thus

$$t_\pi = \frac{\pi}{|A(s_1-s_0)|}$$

for $\Delta m_I=1$. Real gates add refocusing pulses and must account for transverse hyperfine terms and spectators.

3. A complete photon budget

Suppose a node has

$$\begin{aligned} p_{\rm prep}&=0.98,& p_{\rm exc}&=0.95,\\ \beta_{\rm mode}&=0.40,& \eta_{\rm out}&=0.70,\\ \eta_{\rm conv}&=0.35,& \eta_{\rm det}&=0.85. \end{aligned}$$

Ignore other filters. Find the useful detection probability per attempt and identify the largest single multiplicative loss.

Solution

The probability is

$$\begin{aligned} \eta_{\rm node} &=(0.98)(0.95)(0.40) \times(0.70)(0.35)(0.85)\\ &\approx0.0776. \end{aligned}$$

Only about $7.8\%$ of attempts produce a useful detector event. The smallest factor, and therefore the largest single fractional loss here, is the conversion efficiency $0.35$. Improving it does not remove losses in the other factors.

4. Fibre loss and protocol scaling

Two nodes are separated by $25\ \mathrm{km}$ of fibre with attenuation $0.20\ \mathrm{dB/km}$, and detection occurs halfway between them. Find the transmission of each arm and the transmission factors for one detected photon and two required detected photons. Ignore all other losses.

Solution

Each arm is $12.5\ \mathrm{km}$ and has attenuation $2.5\ \mathrm{dB}$:

$$\begin{aligned} \eta_{\rm arm}&=10^{-2.5/10}\approx0.5623,\\ \eta_{\rm arm}^2&\approx0.3162. \end{aligned}$$

The one-photon transmission factor is $0.5623$ and the two-photon factor is $0.3162$, the latter equal to full-length transmission. Using full-length transmission and then squaring it would count twice the traversed distance. Emission, interference, detector, and false-herald factors are still needed to compare actual success rates and fidelities.

5. Waiting time and active memory

A link attempts every $20\ \mu\mathrm s$ with success probability $2\times10^{-4}$ per attempt. The memory coherence under network activity is $0.30\ \mathrm s$. Find the mean link time and active-link efficiency.

Solution

The mean waiting time is

$$\langle T_{\rm link}\rangle = \frac{20\times10^{-6}} {2\times10^{-4}} =0.10\ \mathrm s.$$

Hence

$$\eta_{\rm link}^{\rm active} = \frac{0.30}{0.10}=3.$$

For independent exponential dephasing, $x=1/15000$ and the geometric average gives $\overline C\approx0.749994$. Evaluating at the mean waiting time instead gives $e^{-1/3}\approx0.7165$. If one half of an ideal Bell pair undergoes this dephasing and all other operations are perfect, the final Bell fidelity averaged over successful waiting times is $(1+\overline C)/2\approx0.875$. These are model predictions, not guarantees from the timescale ratio alone.

6. Majority-vote readout

Three independent binary readouts each return the wrong result with probability $p=0.08$. Find the error probability of majority voting. Why can the independence assumption fail for repetitive defect-spin readout?

Solution

A majority is wrong if exactly two or all three outcomes are wrong:

$$\begin{aligned} p_{\rm maj} &=3p^2(1-p)+p^3\\ &=3(0.08)^2(0.92)+(0.08)^3\\ &\approx0.0182. \end{aligned}$$

The nominal error drops from $8\%$ to about $1.8\%$. Repetitions are not independent if optical cycling flips the memory, changes the charge state, or shares a slowly drifting photon rate. In that case a hidden-state model is more appropriate than binomial voting.

7. Useful cavity emission

In the fast-cavity, resonant, negligible-dephasing model, take $C=12$ and $\kappa_{\rm ex}/\kappa=0.70$. What fraction of initial emitter excitations leaves through the useful port? Suppose the uncoupled emitter’s ZPL fraction is $0.03$; should the result be multiplied by $0.03$?

Solution

The cavity branch has probability $12/13$, and the useful-port probability is $(12/13)(0.70)\approx0.6462$. The coupling $g$ entering $C$ is already that of the selected ZPL, so no additional Debye–Waller factor belongs here. Multiplying by $0.03$ again would count the weak ZPL transition twice. Propagation, conversion, and detector losses still multiply this port probability when defined as the corresponding conditional efficiencies.

8. Random phase from optical reset

A nuclear memory acquires a phase $\phi=0.1$ with probability $r=0.05$ during each electron-reset cycle, and otherwise acquires no phase. Cycles are independent. Find the magnitude of the memory coherence after $N=1000$ cycles and distinguish phase tracking from dephasing suppression.

Solution

The off-diagonal density-matrix element is multiplied each cycle by $\lambda=1-r+r e^{-i\phi}$. Hence

$$\begin{aligned} |\lambda|^N&= [1-2r(1-r)(1-\cos\phi)]^{N/2}\\ &\approx0.7887. \end{aligned}$$

An accumulated phase $N\arg\lambda$ can be removed by a known frame update. The magnitude loss remains because the unobserved reset histories have different phases. If the actual phase kick were measured each cycle, one could instead correct that history; a correction of only the average phase does not recover the lost coherence.

9. Heralding versus accepted-state error

A protocol accepts $1\%$ of attempts. Within the accepted ensemble, the probability of a false-herald event is $6\%$ and that of an independent local-gate fault is $4\%$. Find the probability of at least one fault and the probability per attempt of acceptance without either fault. Does this determine the Bell-state fidelity?

Solution

To first order, the probability of at least one fault in an accepted event is

$$\epsilon_{\rm acc}\approx0.06+0.04=0.10.$$

The exact probability of neither fault, given acceptance, is

$$(1-0.06)(1-0.04)=0.9024.$$

Thus the probability per attempt of an accepted state without either error is

$$(0.01)(0.9024)=9.024\times10^{-3}.$$

The distinction matters: the heralding rate is $1\%$, while the useful fault-free accepted-event rate is about $0.902\%$ under this model. The probability of at least one fault is $1-0.9024=0.0976$. These classical labels do not determine the Bell fidelity: a fault branch can retain some target overlap, or two faults can cancel. The branch states are needed before identifying “fault probability” with infidelity.

References

• P.-J. Stas et al., “Robust multi-qubit quantum network node with integrated error detection,” Science 378, 557–560 (2022), doi:10.1126/science.add9771.

• M. H. Abobeih et al., “Fault-tolerant operation of a logical qubit in a diamond quantum processor,” Nature 606, 884–889 (2022), doi:10.1038/s41586-022-04819-6.

• C. P. Anderson et al., “Five-second coherence of a single spin with single-shot readout in silicon carbide,” Science Advances 8, eabm5912 (2022), doi:10.1126/sciadv.abm5912.

• M. Atatüre, D. Englund, N. Vamivakas, S.-Y. Lee, and J. Wrachtrup, “Material platforms for spin-based photonic quantum technologies,” Nature Reviews Materials 3, 38–51 (2018), doi:10.1038/s41578-018-0008-9.

• D. D. Awschalom, R. Hanson, J. Wrachtrup, and B. B. Zhou, “Quantum technologies with optically interfaced solid-state spins,” Nature Photonics 12, 516–527 (2018), doi:10.1038/s41566-018-0232-2.

• S. D. Barrett and P. Kok, “Efficient high-fidelity quantum computation using matter qubits and linear optics,” Physical Review A 71, 060310(R) (2005), doi:10.1103/PhysRevA.71.060310.

• H. Bernien et al., “Heralded entanglement between solid-state qubits separated by three metres,” Nature 497, 86–90 (2013), doi:10.1038/nature12016.

• M. K. Bhaskar et al., “Experimental demonstration of memory-enhanced quantum communication,” Nature 580, 60–64 (2020), doi:10.1038/s41586-020-2103-5.

• C. E. Bradley et al., “A ten-qubit solid-state spin register with quantum memory up to one minute,” Physical Review X 9, 031045 (2019), doi:10.1103/PhysRevX.9.031045.

• A. M. Dibos, M. Raha, C. M. Phenicie, and J. D. Thompson, “Atomic source of single photons in the telecom band,” Physical Review Letters 120, 243601 (2018), doi:10.1103/PhysRevLett.120.243601.

• M. W. Doherty et al., “The nitrogen-vacancy colour centre in diamond,” Physics Reports 528, 1–45 (2013), doi:10.1016/j.physrep.2013.02.001.

• A. L. Falk et al., “Polytype control of spin qubits in silicon carbide,” Nature Communications 4, 1819 (2013), doi:10.1038/ncomms2854.

• S. L. N. Hermans et al., “Qubit teleportation between non-neighbouring nodes in a quantum network,” Nature 605, 663–668 (2022), doi:10.1038/s41586-022-04697-y.

• A. Kiraz, M. Atatüre, and A. İmamoğlu, “Quantum-dot single-photon sources: Prospects for applications in linear optics quantum-information processing,” Physical Review A 69, 032305 (2004), doi:10.1103/PhysRevA.69.032305.

• R. A. Parker et al., “A diamond nanophotonic interface with an optically accessible deterministic electronuclear spin register,” Nature Photonics 18, 156–161 (2024), doi:10.1038/s41566-023-01332-8.

• M. Raha et al., “Optical quantum nondemolition measurement of a single rare earth ion qubit,” Nature Communications 11, 1605 (2020), doi:10.1038/s41467-020-15138-7.

• A. Sipahigil et al., “An integrated diamond nanophotonics platform for quantum-optical networks,” Science 354, 847–850 (2016), doi:10.1126/science.aah6875.

• D. D. Sukachev et al., “Silicon-vacancy spin qubit in diamond: a quantum memory exceeding 10 ms with single-shot state readout,” Physical Review Letters 119, 223602 (2017), doi:10.1103/PhysRevLett.119.223602.

• G. Wolfowicz et al., “Quantum guidelines for solid-state spin defects,” Nature Reviews Materials 6, 906–925 (2021), doi:10.1038/s41578-021-00306-y.